**Title: **Shape Spaces of Curves

**Title: **Complexity in Simple Cross-Sectional
Data with Binary Disease Outcome

**Abstract: **Cros
s-sectionally sampled data with binary disease outcome are commonly collec
ted and analyzed in observational studies for understanding how covariate
s correlate with disease occurrence. At Hopkins SPH and SOM\, cross-sectio
nal data analyses are also commonly included in master and doctoral disse
rtations. This talk will address two questions: (1) Which risk can be ide
ntified in a commonly adopted model (such as the logistic model)? (2) Ar
e there problems when interpreting the identifiable risk? As the progressi
on of a disease typically involves both disease status and duration\, this
talk considers how the binary disease outcome is connected to the progre
ssion of disease through the birth-illness-death process. In general\, we
conclude that the distribution of cross-sectional binary outcome could b
e very different from the population risk distribution. The cross-section
al risk probability is determined jointly by the population risk probabil
ity together with the ratio of duration of diseased state to the duration
of disease-free state. Using the logistic model as an illustrating exam
ple\, we examine the bias from cross-sectional data and argue that the bi
as can almost never be avoided. We present an approach which treats the bi
nary outcome as a specific type of current status data and offers a compro
mised model on the basis of an age-specific risk probability (ARP)\, thou
gh the interpretation of the ARP itself could also be questioned. An anal
ysis based on Alzheimer’s disease data is presented to illustrate the ARP
approach and data complexity. (This is joint work with Yuchen Yang\, Depar
tment of Biostatistics\, Johns Hopkins University).

\n

DTSTART;TZID=America/New_York:20190207T133000 DTEND;TZID=America/New_York:20190207T143000 SEQUENCE:0 SUMMARY:AMS Seminar: Mei Cheng Wang (JHU- Biostat) @ Whitehead 304 URL:https://engineering.jhu.edu/ams/events/ams-seminar-mei-cheng-wang-white head/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-12878@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:

**Title:** Towards Robust and Scalable Private
Data Analysis

**Abstract:**

In the current ag e of big data\, we are constantly creating new data which is analyzed by v arious platforms to improve service and user’s experience. Given the sens itive and confidential nature of these data\, there are obvious security a nd privacy concerns while storing and analyzing such data. In this talk\, I will discuss the fundamental challenges in providing robust security and privacy guarantee while storing and analyzing large data. I will also giv e a brief overview of my contributions and future plans towards addressing these challenges.

\nTo give a glimpse of these challenges in provid ing a robust privacy guarantee known as differential privacy\, I will use spectral sparsification of graphs as an example. Given the ubiquitous natu re of graphs\, differentially private analysis on graphs has gained a lot of interest. However\, existing algorithms for these analyses are tailored made for the task at hand making them infeasible in practice. In this tal k\, I will present a novel differentially private algorithm that outputs a spectral sparsification of the input graph. At the core of this algorithm is a method to privately estimate the importance of an edge in the graph. Prior to this work\, there was no known privacy preserving method that pr ovides such an estimate or spectral sparsification of graphs.

\nSinc e many graph properties are defined by the spectrum of the graph\, this wo rk has many analytical as well as learning theoretic applications. To demo nstrate some applications\, I will show more efficient and accurate analys is of various combinatorial problems on graphs and the first technique to perform privacy preserving manifold learning on graphs.

DTSTART;TZID=America/New_York:20190211T100000 DTEND;TZID=America/New_York:20190211T110000 SEQUENCE:0 SUMMARY:Mathematics Seminar- Jalaj Upadhyay (Computer Science): Optimizatio n and Discrete @ Whitehead 304 URL:https://engineering.jhu.edu/ams/events/mathematics-seminar-optimization -and-discrete-whitehead-304/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-12771@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title:** Optimization and Topology: Two Stori
es

**Abstract: **Algebraic topology and optimization
are typically not considered as closely related fields of mathematics. We
will present two stories of fruitful interaction between these two fields\
, with the implications going the opposite way in the two cases.

I n the first result\, we consider the question of the existence of certain nice decompositions of generalized surfaces called currents in geometric m easure theory. In the finite setting\, we could use tools from algebraic t opology to pose this question as that of the existence of integer solution s to a certain linear programming (LP) problem. Following classical result s on LP that rely on total unimodularity (TU) of matrices\, the answer is known in codimension 1. We develop tools to push this result to the infini te case\, showing that under certain assumptions the TU result from LP imp lies the existence result for codimension 1 currents in general.

\nI n the second story\, we consider new approaches to characterize the robust ness of solutions to a system of nonlinear equations. This problem arises in many applications such as the power grid and other infrastructure netwo rks. We use techniques from algebraic topology (topological degree theory) to characterize the robustness margin of such systems of equations. We th en cast the problem of checking for the specified conditions as a nonlinea r optimization problem. Based on this formulation\, we develop efficient c omputational techniques to estimate lower and upper bounds for the robustn ess margin.

\n\n

DTSTART;TZID=America/New_York:20190214T133000 DTEND;TZID=America/New_York:20190214T143000 SEQUENCE:0 SUMMARY:AMS Seminar: Bala Krishnamorthy (Washington State University) @ Whi tehead 304 URL:https://engineering.jhu.edu/ams/events/ams-seminar-bala-krishnamorthy-w ashington-state-university-whitehead-304/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-12776@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:

**Title:** Mortgage Credit\, Aggregate Demand\,
and Unconventional Monetary Policy

**Abstract: **I d
evelop a quantitative model of the mortgage market operating in an economy
with financial frictions and nominal rigidities. I use this model to stud
y the effectiveness of large-scale asset purchases (LSAPs) by a central ba
nk as a tool of monetary policy. When negative shocks hit\, homeowner and
financial sector balance sheets are impaired\, borrowing constraints bind\
, asset prices and aggregate demand drop\, hampering the transmission of c
onventional monetary policy. LSAPs boost aggregate demand in a crisis by d
irecting additional lending to homeowners\, raising house prices\, and sta
blishing expectations of future financial stability. However\, legacy hous
ehold debt depresses output and consumption in recovery. In the long run\,
a commitment to ongoing use of LSAPs in crises reduces credit and busines
s cycle volatility and redistributes resources from borrowers and intermed
iaries to savers.

**Title:** Big Data is Low Rank

**Abstract: **Matrices of low rank are pervasive in big data\, app
earing in recommender systems\, movie preferences\, topic models\, medical
records\, and genomics.

While there is a vast literature on how t o exploit low rank structure in these datasets\, there is less attention o n explaining why low rank structure appears in the first place.

\nIn this talk\, we explain the abundance of low rank matrices in big data by proving that certain latent variable models associated to piecewise analyt ic functions are of log-rank. Any large matrix from such a latent variable model can be approximated\, up to a small error\, by a low rank matrix.\n

Armed with this theorem\, we show how to use a low rank modeling fr amework to exploit low rank structure even for datasets that are not numer ic\, with applications in the social sciences\, medicine\, retail\, and ma chine learning.

DTSTART;TZID=America/New_York:20190228T133000 DTEND;TZID=America/New_York:20190228T143000 SEQUENCE:0 SUMMARY:AMS Seminar: Madeleine Udell (Cornell University) @ Whitehead 304 URL:https://engineering.jhu.edu/ams/events/ams-seminar-madeline-udell-corne ll-university-whitehead-304/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-12781@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title: **Market Microstructure Invariance: A
Dynamic Equilibrium Model

**Abstract: **Invariance re
lationships are derived in a dynamic\, infinite-horizon\, equilibrium mode
l of adverse selection with risk-neutral informed traders\, noise traders\
, risk-neutral market makers\, and endogenous information production. Scal
ing laws for bet size and transaction costs require the assumption that th
e effort required to generate one bet does not vary across securities and
time. Scaling laws for pricing accuracy and market resiliency require the
additional assumption that private information has the same signal-to-nois
e ratio across markets. Prices follow a martingale with endogenously deriv
ed stochastic volatility. Returns volatility\, pricing accuracy\, market d
epth\, and market resiliency are closely related to one another. The model
solution depends on two state variables: stock price and hard-to- observe
pricing accuracy. Invariance makes predictions operational by expressing
them in terms of log-linear functions of easily observable variables such
as price\, volume\, and volatility.

DTSTART;TZID=America/New_York:20190307T133000 DTEND;TZID=America/New_York:20190307T143000 SEQUENCE:0 SUMMARY:AMS Seminar: Albert “Pete” Kyle (University of MD) @ Whitehead 304 URL:https://engineering.jhu.edu/ams/events/ams-seminar-albert-pete-kyle-uni versity-of-md-whitehead-304/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-12786@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:

**Title: **Uncertainty Quantification and Nonpa
rametric Inference for Complex Data and Simulations

**Abstra
ct: **Recent technological advances have led to a rapid growth in
not just the amount of scientific data but also their complexity and richn
ess. Simulation models have\, at the same time\, become increasingly detai
led and better at capturing the underlying processes that generate observa
ble data. On the statistical methods front\, however\, we still lack tools
that accurately quantify complex relationships between data and model par
ameters\, as well as adequate tools to validate models of multivariate lik
elihoods and posteriors. In this talk\, I will discuss our current work on
addressing some of the multi-faceted challenges encountered in astronomy
but more generally applicable to fields involving massive amounts of compl
ex data and simulations\; in particularly\, challenges related to (i) buil
ding conditional probability models that can handle inputs of different mo
dalities\, e.g. photometric data and correlation functions\, (ii) estimati
ng non-Gaussian likelihoods and posteriors via simulations\, and (iii) ass
essing the performance of complex models and simulations when the true dis
tributions are not known. I will draw examples from photometric redshift e
stimation and from the inference of cosmological parameters. (Part of this
work is joint with Rafael Izbicki\, Taylor Pospisil\, Peter Freeman\, Ilm
un Kim\, and the LSST-DESC PZ working group)

**Title:** Guiding clinical and preclinical inv
estigations of breast cancer with mathematical modeling and analyses

**Abstract**: One of the great challenges for cancer treat
ment is the inability to optimize therapy. Without a reasonable mathematic
al framework\, our ability to select treatment regimens for the individual
patient is fundamentally limited to trial and error. Presented here are e
xamples of data-driven\, integrated experimental-mathematical approaches t
o studying breast cancer’s response to therapy for both pre-clinical and c
linical investigations. The preclinical model\, consisting of ODEs\, conne
cts various experiments for an *in vivo* mouse system to better und
erstand the interactions of the immune response and targeted therapy for b
reast cancer. The clinical model is a 3D PDE system for predicting tumor r
esponse to neoadjuvant therapy using patient-specific data that lays the g
roundwork for optimizing chemotherapeutic dosing and scheduling. In both e
xamples\, the results of uncertainty and sensitivity analyses are discusse
d to show how they can be used to generate experimentally testable hypothe
ses\, narrow the scope for experimental investigations\, and evolve mathem
atical models. Additionally\, multi-scale models are proposed that bridge
the gap between *in vitro* and *in vivo* experiments to step
towards clinical translation.

**Title: **
Min-Max Relations for Packing and Covering

**Abstract: We consider a family M of subsets of a finite set E. A “cover” is a su
bset of E that intersects every member of the family M. A “packing” is a s
et of members of M no two of which intersect. Clearly\, the cardinality of
a packing is at most that of a cover. We study conditions under which the
maximum cardinality of a packing equals the minimum cardinality of a cove
r. We present recent results obtained jointly with Ahmad Abdi and Dabeen L
ee.**

**Bio: ** Gerard Cornuejols is professor of Opera
tions Research at Carnegie Mellon University. His research interests are i
n integer programming and combinatorial optimization. He received the Lanc
hester Prize twice (1978 and 2015)\, the Fulkerson Prize (2000)\, the Dant
zig Prize (2009) and the von Neumann Theory Prize (2011).

**Title: **Robust inference with the knockoff f
ilter.

**Abstract: **In this talk\, I will present on
going work on the knockoff filter for inference in regression. In a high-d
imensional model selection problem\, we would like to select relevant feat
ures without too many false positives. The knockoff filter provides a tool
for model selection by creating knockoff copies of each feature\, testing
the model selection algorithm for its ability to distinguish true from fa
lse covariates to control the false positives. In practice\, the modeling
assumptions that underlie the construction of the knockoffs may be violate
d\, as we cannot know the exact dependence structure between the various f
eatures. Our ongoing work aims to determine and improve the robustness pro
perties of the knockoff framework in this setting. We find that when knock
off features are constructed using estimated feature distributions whose e
rrors are small in a KL divergence type measure\, the knockoff filter prov
ably controls the false discovery rate at only a slightly higher level. Th
is work is joint with Emmanuel Candès and Richard Samworth.

This is joint wor k with Emmanuel Candès\, Aaditya Ramdas\, and Ryan Tibshirani.

\n**Bio: **TBA

**Title:** Distribution free prediction: Is conditional in
ference possible?

**Abstract:** We consider the problem o
f distribution-free predictive inference\, with the goal of producing pred
ictive coverage guarantees that hold conditionally rather than marginally.
Existing methods such as conformal prediction offer marginal coverage gua
rantees\, where predictive coverage holds on average over all possible tes
t points\, but this is not sufficient for many practical applications wher
e we would like to know that our predictions are valid for a given individ
ual\, not merely on average over a population. On the other hand\, exact c
onditional inference guarantees are known to be impossible without imposin
g assumptions on the underlying distribution. In this work we aim to explo
re the space in between these two\, and examine what types of relaxations
of the conditional coverage property would alleviate some of the practical
concerns with marginal coverage guarantees while still being possible to
achieve in a distribution-free setting.

This is joint work with Emmanu el Candès\, Aaditya Ramdas\, and Ryan Tibshirani.

\n**B
io: **TBA

**Title: **“Real-tim
e” optimization under forward rank-dependent processes: time-consistent op
timality under probability distortions

**Abstract: Forward performance processes are define
d via time-consistent optimality and incorporate “real-time” incoming info
rmation. On the other hand\, popular performance criteria – for example\,
mean-variance optimization\, hyperbolic discounting\, probability distorti
ons – are by nature time-inconsistent. How to define forward performance c
riteria in time-inconsistent settings then becomes a challenging problem\,
both conceptually and technically. In this talk\, I will discuss the case
of probability distortions and introduce the concept of forward rank-depe
ndent performance processes. Among others\, I will show how forward probab
ility distortions are affected by “real-time” changes in the stochastic en
vironment and\, also\, present a striking equivalence between forward rank
-dependent criteria and time-monotone forward processes under appropriate
measure-changes. A byproduct of the work is a novel result on the so-calle
d dynamic utilities and on time-inconsistent problems in the classical (ba
ckward) setting. **

**Title: **Uncertainty propagation in mechanics
and materials by design based on surrogate model development

** Abstract: **With the onset of advanced manufacturing capabiliti
es and in situ characterization techniques\, simultaneous material/structu
ral design is becoming increasingly feasible for maximum structural perfor
mance. At the heart of such design processes is the availability of multi-
scale mechanics models that incorporate explicit representation of the mat
erial (such as microstructural descriptors) and the structure (such as the
geometry). A major challenge here is that a full physically-based multi-s
cale model is often computationally infeasible. Surrogate functions that p
rovide a simplified representation of the material provide a much more eff
icient alternative. Such surrogate functions also enable a quantification
of the propagation of uncertainties between scales. While these surrogate
functions do increase efficiency\, they lead to a number of associated cha
llenges. If the material is represented by a large number of microstructur
al parameters\, then the high dimensionality of the surrogate function req
uires many samples in order to build an accurate surrogate. Furthermore\,
some micro-scale behavior\, such as sudden damage\, can lead to discontinu
ities in the surrogate function\, which makes it difficult to interpolate
or collocate the results. This seminar will describe a number of approache
s to building surrogates\, including cases in which the micro-scale model
provides key response values and/or gradients of key response values.

~~Title: ~~~~Modeling Particulate Air Pollution for Inference Abo
ut Neurodegenerative Effects~~

**Abstract: **Evidence i
s accumulating to support a link between chronic air pollution exposures a
nd neurotoxic effects. For instance\, EPA’s most recent Integrated Scienc
e Assessment for particulate matter concluded that the associations betwee
n PM_{2.5} and nervous system effects\, including brain inflammati
on\, oxidative stress\, reduced cognitive function\, and neurodegeneration
\, are likely causal. We are conducting an epidemiologic cohort study\, t
he Adult Changes in Thought Air Pollution (ACT-AP) study\, to determine wh
ether\, in an elderly population free of dementia at baseline\, long-term
air pollution exposure is associated with cognitive decline\, incidence of
Alzheimer’s disease and all-cause dementia\, and adverse neuruopathologic
al changes in brain tissue. For exposure assessment in this study we are
modeling criteria air pollutants using existing regulatory monitoring data
supplemented with measurements from low-cost sensors. One important scie
ntific question we are addressing is whether low-cost sensor data improve
our ability to quantify PM_{2.5 }exposure in the Puget Sound. I w
ill discuss our approach and our preliminary conclusions that suggest that
low-cost sensor can improve exposure assessment in epidemiologic cohort s
tudies. I will also describe the innovative mobile monitoring campaign we
have just started. We designed this campaign with epidemiologic inferen
ce in mind\; it will allow us to estimate whether there are adverse effect
s to the brain associated with infrequently monitored traffic-related poll
utants\, including ultrafine particles and black carbon.

**B
io: **Dr. Sheppard is Professor and Assistant Chair of Environmenta
l and Occupational Health Sciences and Professor of Biostatistics. Her cur
rent research portfolio includes several studies of air pollution exposure
s and their neurotoxicant effects. She has a Ph.D. in biostatistics. Her m
ethodologic interests center on observational study methods\, exposure mod
eling\, and epidemiology\, and\; her applied research focuses on the the h
ealth effects of occupational and environmental exposures. She is principa
l investigator of a NIH-funded training grant called Biostatistics\, Epide
miologic & Bioinformatics Training in Environmental Health and SURE-EH\, a
project to promote diversity in the environmental health sciences. She le
ads the biostatistical cores for several projects and collaborates with DE
OHS faculty on air pollution cohort studies\, identifying the effects of m
ultipollutant exposures\, and studying manganese exposures. She is a membe
r of the Epidemiology editorial board\, the Health Effects Institute Revie
w Committee\, the EPA Clean Air Scientific Advisory Committee \, and has s
erved on the several EPA Scientific Advisory Panels\, most recently for th
e Carcinogenic Potential of Glyphosate. Board Chemical Assessment Advisory
Committees for Ethylene Oxide Review and for Toxicological Review of Libb
y Amphibole Asbestos.

**Title: **Enter the matrix: interpreting biolo
gical systems through matrix factorization and transfer learning of single
cell data

**Abstract: **Next generation and single c
ell sequencing have ushered in an era of big data in biology. These data
present an unprecedented opportunity to learn new mechanisms and ask unask
ed questions. Matrix factorization (MF) techniques can reveal low-dimensi
onal structure from high-dimensional data to uncover new biological knowle
dge. The knowledge of gained from low dimensional features in training da
ta can also be transferred to new datasets to relate disparate model syste
ms and data modalities. We illustrate the power of these techniques for i
nterpretation of high dimensional data through case studies in postmortem
tissues from GTEx\, acquired therapeutic resistance in cancer\, and develo
pmental biology.

AMS New Student Picnic will be held in Great Hall at 12pm.\n

Come out to meet and greet other incoming graduate students and you r professors.

DTSTART;TZID=America/New_York:20190828T120000 DTEND;TZID=America/New_York:20190828T133000 SEQUENCE:0 SUMMARY:AMS New Graduate Student Picnic- Great Hall @ 12pm URL:https://engineering.jhu.edu/ams/events/ams-new-student-picnic-mattin-ro oftop-12pm/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-13629@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title: TBA**

\n

**Abstr
act: TBA**

**Title: **Diffeomorphic Learning

**Title: **Mobius Registration

**
Abstract: **Conformal spherical parametrizations of genus-zero surf
aces have been explored as a way to represent surfaces over a canonical do
main. This\, in turn\, makes it possible to establish correspondences betw
een pairs of shapes\, enabling applications like interpolation and detail
transfer. However\, one challenge of using these parametrizations is that
they are only unique up to Mobius transformation. As such\, to use these p
arametrization for establishing correspondences between shapes\, it is fir
st necessary to register the two spherical parametrizations with respect t
o the Mobius transformations. That is\, to find the Mobius transformation
best aligning the two spherical maps.

In this talk we will address the problem of Mobius registration by expresing the space of Mobius trans formations as the composition of inversions and rotations. We will show th at these two classes of transformation are fundamentally different: Each s pherical parametrization can be canonically normalized to remove inversion ambiguity. And\, using existing techniques for fast correlation over SO(3 )\, pairs of spherical parametrizations can be rotationally aligned. We wi ll consider implications of inversion normalization for the calculation of spherical orbifolds and will conclude by discussing why Mobius registrati on may not be sufficient.

DTSTART;TZID=America/New_York:20190926T133000 DTEND;TZID=America/New_York:20190926T143000 SEQUENCE:0 SUMMARY:AMS Seminar: Misha Kazhdan (JHU- Computer Science) @ Whitehead 304 URL:https://engineering.jhu.edu/ams/events/ams-seminar-tba-whitehead-304-11 / X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-13640@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title**: Multivariate Pareto Records

**Abstract: **Consider i.i.d. *d*-dimensional observat
ions with independent coordinates\, each with (say) the standard Exponenti
al distribution. Say that the *n*-th observation *sets a (Paret
o) record* if it is not dominated by any of the first *n* – 1 o
bservations. If *k* is in {1\, …\, *n*}\, say that the *
k*-th observation is a *current record* at time *n* if i
t sets a record and is not dominated by any of the next *n* – *k
* observations\; and say that the *n*-th observation *breaks
the record* set by the *k*-th observation if the *k*-th
observation is a current record at time *n* – 1 but not at time

We will discuss one or more of the following topics: (i) an efficient algorithm for the simulation of Pareto records\, and its (pa rtial) analysis\; (ii) the location and thickness of the record frontier\; (iii) how the Geometric(1/2) distribution arises in connection with the b reaking of bivariate records.

\nThis is joint work with Daniel Q. Na iman.

DTSTART;TZID=America/New_York:20191003T133000 DTEND;TZID=America/New_York:20191003T143000 SEQUENCE:0 SUMMARY:AMS Seminar: Jim Fill (JHU- AMS) @ Whitehead 304 URL:https://engineering.jhu.edu/ams/events/ams-seminar-tba-whitehead-304-12 / X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-13619@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:\n**Title:**
Online Experimentation and Learning Algorithms in a Clinical Trial

**Abstract:** In this talk we describe two reinforcement le
arning algorithms we have implemented in a mobile health physical activity
trial. These algorithms are designed to tackle two challenges faced by mo
bile health. The first challenge is that while most treatments delivered b
y a mobile device have immediate nonnegative (hopefully positive) effects\
, longer term effects tend to be negative due to user burden. To address t
his first challenge we add a low variance proxy for the delay effects to t
he reward (e.g. immediate response) in the learning algorithm. The second
challenge is that data on any one individual is very noisy making it diffi
cult for the algorithm to learn. To address this challenge we pool data ac
ross participants.

**Bio: **Susan Murphy is Professor
of Statistics at Harvard University\, Radcliffe Alumnae Professor at the
Radcliffe Institute\, Harvard University\, and Professor of Computer Scien
ce at the Harvard John A. Paulson School of Engineering and Applied Scienc
es. Her lab works on clinical trial designs and learning algorithms for d
eveloping mobile health policies. She is a 2013 MacArthur Fellow\, a mem
ber of the National Academy of Sciences and the National Academy of Medici
ne\, both of the US National Academies. She is currently President of the
Institute of Mathematical Statistics.

Susan Murphy’s website is < a href='http://people.seas.harvard.edu/~samurphy/'>http://people.seas.harv ard.edu/~samurphy/

DTSTART;TZID=America/New_York:20191010T133000 DTEND;TZID=America/New_York:20191010T143000 SEQUENCE:0 SUMMARY:The Acheson J. Duncan Lecture Series: AMS Seminar: Susan Murphy (Ha rvard University) @ Maryland 110 URL:https://engineering.jhu.edu/ams/events/the-acheson-j-duncan-lecture-ser ies-ams-seminar-susan-murphy-harvard-university-maryland-110/ X-COST-TYPE:free X-WP-IMAGES-URL:thumbnail\;https://engineering.jhu.edu/ams/wp-content/uploa ds/2019/08/Susan-Murphy-200x300.jpg\;200\;300\,medium\;https://engineering .jhu.edu/ams/wp-content/uploads/2019/08/Susan-Murphy-200x300.jpg\;200\;300 \,large\;https://engineering.jhu.edu/ams/wp-content/uploads/2019/08/Susan- Murphy-200x300.jpg\;200\;300\,full\;https://engineering.jhu.edu/ams/wp-con tent/uploads/2019/08/Susan-Murphy-200x300.jpg\;200\;300 END:VEVENT BEGIN:VEVENT UID:ai1ec-13644@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title: **Latent variable models for biomarker
s of Alzheimer’s disease

**Abstract: **Accumulating
evidence suggests that the initiation of Alzheimer’s disease (AD) pathogen
ic process precedes the first symptoms by a decade or more. The recognitio
n of this decade-long asymptomatic stage has greatly impact AD research an
d therapeutic development to focus on the preclinical stage of AD pathogen
ic process\, at which time disease-modifying therapy is more likely to be
effective. On the other hand\, the decade-long preclinical stage imposes a
major challenge in investigating biomarkers for early AD detection. The c
hallenge is two-fold. Firstly\, the unobservable disease status leads to c
hallenges in evaluating the potential diagnostic capacity of AD biomarkers
. Clinical diagnoses are often used in current evaluation\, but they are k
nown to be error-prone\, especially in the early course of the disease. In
addition\, the clinical diagnosis is often based on or provide an unfair
advantage to current standard tests. Therefore\, they can mask the prognos
tic value of a useful biomarker\, especially when the biomarker is much mo
re accurate than the standard tests. Since AD pathophysiology has been rec
ognized as a multidimensional process that involves amyloid deposition\, n
eurofibrillary tangles\, and neurodegeneration among other aspects\, we pr
oposed a latent variable model to study the underlying AD pathophysiology
process revealed by multidimensional markers. Secondly\, the unobservable
disease process also leads to challenges in understanding the ordering and
shape of AD biomarker cascade\, which is critical for early detection and
therapeutic development yet is still under great debate. I will present a
work-in-progress that attempts to inform this debate and outline a model
that considers continuous latent disease progression.

**Title: **Optimal Oil Production and Taxation
under Carbon Emission Constraints.

**Abstract: **We
study the optimal extraction policy of an oil field as well as the efficie
nt taxation of the revenues generated in light of various economic restric
tions and constraints. Taking into account the fact that the oil price in
worldwide commodity markets fluctuates randomly following global and seaso
nal macroeconomic parameters\, we model the evolution of the oil price as
a mean reverting regime-switching jump diffusion process. Moreover\, takin
g into account the fact that oil producing countries rely on oil sale reve
nues as well as taxes levied on oil companies for a good portion of the re
venue side of their budgets\, we formulate this problem as a differential
game where the two players are the mining company whose aim is to maximize
the revenues generated from its extracting activities and the government
agency in charge of regulating and taxing natural resources. We prove the
existence of a Nash equilibrium and characterize the value functions of th
is stochastic differential game as the unique viscosity solutions of the c
orresponding Hamilton Jacobi Isaacs equations. Furthermore\, optimal extra
ction and fiscal policies that should be applied when the equilibrium is r
eached are derived. A numerical example is presented to illustrate these r
esults.

**Title: **Optimal Transport-Based Distances f
or Metric Space Matching

**Abstract: **I will overvie
w some methods for comparing datasets modeled as metric measure spaces (mm
-spaces)\, which are compact metric spaces endowed with probability measur
es. The main tool is Gromov-Wasserstein (GW) distance\, which provides a m
etric on the collection of all mm-spaces. The definition of GW distance is
inspired by ideas from optimal transport and it has fascinating connectio
ns to many other areas of mathematics. I will discuss theoretical results
on estimating GW distance using distribution-valued invariants of mm-space
s as well as some work on the use of GW distance for practical application
s in data science.

**Title: **Optimal Singular Value Decompositio
n for High-dimensional High-order Data

**Abstract: **High-dimensional high-order data arise in many modern scientific applicat
ions including genomics\, brain imaging\, and social science. In this talk
\, we consider the methods\, theories\, and computations for tensor singul
ar value decomposition (tensor SVD)\, which aims to extract the hidden low
-rank structure from high-dimensional high-order data. First\, comprehensi
ve results are developed on both the statistical and computational limits
for tensor SVD under the general scenario. This problem exhibits three dif
ferent phases according to signal-noise-ratio (SNR)\, and the minimax-opti
mal statistical and/or computational results are developed in each of the
regimes. In addition\, we consider the sparse tensor singular value decomp
osition which allows more robust estimation under sparsity structural assu
mptions. A novel sparse tensor alternating thresholding algorithm is propo
sed. Both the optimal theoretical results and numerical analyses are provi
ded to guarantee the performance of the proposed procedure.

DTSTART;TZID=America/New_York:20191031T133000 DTEND;TZID=America/New_York:20191031T143000 SEQUENCE:0 SUMMARY:AMS Seminar: Anru Zhang (University of Wisconsin-Madison) @ Whitehe ad 304 URL:https://engineering.jhu.edu/ams/events/ams-seminar-tba-whitehead-304-15 / X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-13655@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:

**Title: **Computational methods for Quantifyin
g Gerrymandering and other computational statistical mechanics problems.\n

**Abstract: **I will describe some of the interesting
problems which have arisen around the problem of understanding Gerrymander
ing. It is a high sampling dimensional problem. I will talk about some bas
ic MCMC schemes and some extensions to both interesting global moves as we
ll as some other generalizations. I will also take a moment to frame the p
roblem and state some open questions

DTSTART;TZID=America/New_York:20191107T133000 DTEND;TZID=America/New_York:20191107T143000 SEQUENCE:0 SUMMARY:AMS Seminar: Jonathan Mattingly (Duke University) @ Whitehead 304 URL:https://engineering.jhu.edu/ams/events/ams-seminar-tba-whitehead-304-16 / X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-14614@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:\n

**Title:**
Matroids and Optimum Branching Systems

**Abstract:**
The Optimum Branching Systems problem (OBS) is Given a directed graph G\,
specified root nodes r(i)\, and a cost for each edge of G\, find a least c
ost collection of edge-disjoint directed spanning trees in G\, rooted resp
ectively at the nodes r(i)\, i.e.\, r(i)-branchings in G. We describe a po
lynomial time algorithm for OBS. The mincost network flow problem is a spe
cial case of OBS. However OBS does not reduce to it. By letting M1 and M2
be certain matroids\, matroid intersection solves OBS\, and a bunch of oth
er combinatorial optimization problems. Matroid intersection is the only a
pproach known for solving OBS. Curiously\, the simplest way known to descr
ibe an algorithm for OBS is by matroid intersection for general matroids.
Matroid algorithms use subroutine sources of the matroids\, M\, which say
when a set is independent in M. The ingredients for solving OBS are in boo
ks on combinatorial optimization\, but I’m still trying get people to do a
good computer implementation. It’s clearly possible\, but a bit complicat
ed.

**Bio: **Jack Edmonds is a John von Neumann Theor
y Prize recipient and one of the creators of the field of combinatorial op
timization and polyhedral combinatorics. His 1965 paper “*Paths\, Trees
and Flowers” *was one of the first papers to suggest the possibility
of establishing a mathematical theory of efficient combinatorial algorithm
s. In that paper and in the subsequent paper “*Maximum Matching and a <
/em> Polyhedron with 0-1 Vertices” Edmonds gave remarkable polynom
ial-time algorithms for the construction of maximum matchings. Even more i
mportantly these papers showed how a good characterization of the polyhedr
on associated with a combinatorial optimization problem could lead\, via t
he duality theory of linear programming\, to the construction of an effici
ent algorithm for the solution of that problem. In 2014 he was honored as
a Distinguished Scientist and inducted into the National Institute of Stan
dards and Technology’s Gallery for his “fundamental contributions in c
ombinatorial optimization\, discrete mathematics\, and the theory
of computing.”*

\n

edmonds_goldman – **slides**

**Title: **Matrix Means and a Novel High-Dimens
ional Shrinkage Phenomenon

**Abstract: **Many statist
ical settings call for estimating a population parameter\, most typically
the population mean\, from a sample of matrices. The most natural estimate
of the population mean is the arithmetic mean\, but there are many other
matrix means that may behave differently\, especially in high dimensions.
Here we consider the matrix harmonic mean as an alternative to the arithme
tic matrix mean. We show that in certain high-dimensional regimes\, the ha
rmonic mean yields an improvement over the arithmetic mean in estimation e
rror as measured by the operator norm. Counter-intuitively\, studying the
asymptotic behavior of these two matrix means in a spiked covariance estim
ation problem\, we find that this improvement in operator norm error does
not imply better recovery of the leading eigenvector. We also show that a
Rao-Blackwellized version of the harmonic mean is equivalent to a linear s
hrinkage estimator that has been studied previously in the high-dimensiona
l covariance estimation literature. Simulations complement the theoretical
results\, illustrating the conditions under which the harmonic matrix mea
n yields an empirically better estimate.

**Title: **Spatial-temporal modeling of mechanoch
emistry of cellular processes

**Abstract: **The centr
al theme of Dr. Jian Liu’s research is to understand how mechanical action
s feedback to biochemical pathways in cellular processes\, and how such me
chanochemical crosstalk among key cellular players governs spatial-tempora
l regulation and shapes cell functions. He confronts these challenges by b
ringing together theoretical and computational studies\, rooted in statist
ical mechanics\, with a diversity of biological experiments. His seminar w
ill provide an overview of his current research interest – spanning the fi
elds of cell migration\, cell division\, and membrane trafficking – with a
particular emphasis on membrane shape-mediated excitability in cellular p
rocesses.

**Biography**: Dr. Jian Liu graduated from
Peking University with a B.S. in chemistry in 2000 and earned his Ph.D. in
theoretical chemistry from the University of California\, Berkeley in 200
5. He completed postdoctoral fellowships at the University of California\,
San Diego\, Center for Theoretical Biological Physics from 2005 to 2007 a
nd at the University of California\, Berkeley\, Department of Molecular an
d Cell Biology in the laboratory of George Oster from 2007 to 2009. Dr. Li
u joined the NHLBI as a principal Investigator in 2010. Dr. Liu takes a di
stinct approach to theoretical biology\, treating cellular systems as disc
rete functional modules comprising a set of critical players. This allows
both simplification and retention of essential biological features. The la
rger goal of this modular approach is to allow for processes to be combine
d at a theoretical level to reveal the interplay among them in the cell as
a whole.

\n

\n

DTSTART;TZID=America/New_York:20191121T133000 DTEND;TZID=America/New_York:20191121T143000 SEQUENCE:0 SUMMARY:AMS Seminar: Jian Liu (JHMI) @ Whitehead 304 URL:https://engineering.jhu.edu/ams/events/ams-seminar-tba-whitehead-304-18 / X-COST-TYPE:free X-WP-IMAGES-URL:thumbnail\;https://engineering.jhu.edu/ams/wp-content/uploa ds/2019/08/jian-liu-214x300.jpg\;214\;300\,medium\;https://engineering.jhu .edu/ams/wp-content/uploads/2019/08/jian-liu-214x300.jpg\;214\;300\,large\ ;https://engineering.jhu.edu/ams/wp-content/uploads/2019/08/jian-liu-214x3 00.jpg\;214\;300\,full\;https://engineering.jhu.edu/ams/wp-content/uploads /2019/08/jian-liu-214x300.jpg\;214\;300 END:VEVENT BEGIN:VEVENT UID:ai1ec-13661@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:

**Title: **Wave propagation in inhomogeneous me
dia: An introduction to Generalized Plane Waves

**Abstract:
**Trefftz methods rely\, in broad terms\, on the idea of approximat
ing solutions to PDEs using basis functions which are exact solutions of t
he Partial Differential Equation (PDE)\, making explicit use of informatio
n about the ambient medium. But wave propagation problems in inhomogeneous
media is modeled by PDEs with variable coefficients\, and in general no e
xact solutions are available. Generalized Plane Waves (GPWs) are functions
that have been introduced\, in the case of the Helmholtz equation with va
riable coefficients\, to address this problem: they are not exact solution
s to the PDE but are instead constructed locally as high order approximate
solutions. We will discuss the origin\, the construction\, and the proper
ties of GPWs. The construction process introduces a consistency error\, re
quiring a specific analysis.

**Title:** “Measures in geometry: a look at two
cases of fruitful interaction.”

**Abstract:** “This
talk will explore the use of measures as a convenient way to represent and
analyze the shape of objects\, mathematically and/or numerically. We will
specifically focus on two particular examples of such interactions.

In the first part of the talk\, I will introduce\, in a simple setting\ , the so called length measures associated to planar closed curves. Althou gh the length measure do not fully characterize the underlying shape\, the celebrated Minkowski-Fenchel-Jessen theorem shows that this is the case w hen restricting to the subclass of convex shapes. This equivalence has bee n key to the derivation of many important results in the field of convex g eometry. We will mention some of these such as isoperimetric inequalities and discuss several open questions related to length measures for non-conv ex shapes.

\nIn the second part of the talk\, I will present another class of measures called varifolds\, which allow this time to represent i njectively any shape such as curves\, surfaces or even submanifolds of any dimension. I will then examine the construction of numerically tractable metrics based on varifolds which can be used to formulate and tackle vario us problems in shape analysis. We will focus in particular on the problems of compression (or quantization) of varifolds and of diffeomorphic regist ration between two shapes.”

DTSTART;TZID=America/New_York:20200130T133000 DTEND;TZID=America/New_York:20200130T143000 SEQUENCE:0 SUMMARY:AMS Seminar: Nicolas Charon (JHU- Applied Math & Stats) @ Whitehead 304 URL:https://engineering.jhu.edu/ams/events/ams-seminar-nicolas-charon-jhu-a pplied-math-stats-whitehead-304/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-14883@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title:** Cancer is a disease of epigenetic st
ochasticity

**Abstract**: I proposed in 2006 (Nat Rev
Genet) that increased epigenetic stochasticity is a driving force of tumo
r progression from its origin to metastasis\, and would allow rapid select
ion for tumor cell survival at the expense of the host. This model puts ep
igenetic instability at the heart of tumor progression and is the primary
target of cancer mutations. Several recent observations from the laborator
y confirm the model\, and establish mechanisms including disruption of the
epigenome that involving blocks of DNA hypomethylation and heterochromati
n\, and metabolic changes involving the oxidative branch of the pentose ph
osphate pathway. We have recently developed mathematically rigorous Gibbs-
Boltzmann-style epigenetic landscapes incorporating stochasticity and show
n its relationship to entropy in information theory. Recent data shows tha
t this approach identifies epigenetic *and genetic* drivers of canc
er\, using acute lymphoblastic leukemia as a model\, as well as the close
relationship between entropy in cancer and entropy in stem cell reprogramm
ing.

DTSTART;TZID=America/New_York:20200206T133000 DTEND;TZID=America/New_York:20200206T143000 SEQUENCE:0 SUMMARY:AMS Weekly Seminar: Andy Feinberg (Bloomberg Distinguished Professo r\, Schools of Medicine\, Engineering\, and Public Health\, Johns Hopkins University) @ Whitehead 304) URL:https://engineering.jhu.edu/ams/events/ams-weekly-seminar-andy-feinberg -jhu-som-dom-molecular-medicine/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-14972@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:

**Title:** Inverse semigroup theory of cutting
planes for integer linear optimization

**Abstract:**
MIP practitioners can solve large-scale mixed integer optimizationproblems
to optimality or near-optimality by competent modelizationand use of bran
ch-and-cut solvers. This technology was enabled to alarge part by the rev
ival of Gomory’s classic general-purpose cuttingplanes such as the Gomory
mixed integer cut. In the theory of such general-purpose cutting planes (v
alidinequalities) the traditional\, finite-dimensional techniques ofpolyhe
dral combinatorics are complemented by infinite-dimensionalmethods\, the s
tudy of cut-generating functions. In my talk I will introduce the classic
Gomory-Johnson model\, auniversal relaxation of integer programs in the fo
rm of a singleconstraint in infinitely many nonnegative integer variables.
Thenondominated valid inequalities (cut-generating functions) for thismo
del\, “minimal functions”\, are characterized by functionalinequalities su
ch as subadditivity. Given a minimal function\, we are interested in findi
ng improvingdirections that lead to stronger cuts and eventually to “extre
mefunctions”\, which cannot be strengthened further — an analogue offacet-
defining inequalities. I will present an inverse semigroup theory for mini
mal functions\,which enables us to obtain a complete description of the sp
ace of”improving directions” (perturbations) of a minimal function. This
isjoint work with Robert Hildebrand and Yuan Zhou\, which appeared inIPCO
2019\; a full paper is available at https://arxiv.org/abs/1811.06189 .

**Title: **Edge-Selection Priors for Graphical
Models and Applications to Complex Biological Data

**Abstrac
t: **There is now a huge literature on Bayesian methods for variabl
e selection in linear models that use spike-and-slab priors.

Such methods\, in particular\, have been quite successful for applications in a variety of different fields. A parallel methodological development has ha ppened in graphical models\, where priors are specified on precision matri ces. In this talk I will describe priors for edge selection for the estima tion of multiple graphs that may share common features\, such as presence/ absence of edges or strengths of connections. I will also describe modelin g frameworks for non-Gaussian data and discuss computational challenges.\n

I will motivate the development of the models using specific applic ations from neuroimaging and from studies that use biomedical data.

DTSTART;TZID=America/New_York:20200213T133000 DTEND;TZID=America/New_York:20200213T143000 SEQUENCE:0 SUMMARY:AMS Seminar: Dr. Marina Vannucci (Rice University) @ Whitehead 304 URL:https://engineering.jhu.edu/ams/events/ams-seminar-marina-vannucci-rice -university-whitehead-304/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-14931@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title: **A scaling-invariant algorithm for li
near programming whose running time depends only on the constraint matrix<
/p>\n

**Abstract: **Following the breakthrough work of Tard
os in the bit-complexity model\, Vavasis and Ye gave the first exact algor
ithm for linear programming in the real model of computation with running
time depending only on the constraint matrix. For solving a linear program
(LP) max cx\,Ax=b\,x>=0\,A in R^{mxn}\, Vavasis and Ye developed a primal
-dual interior point method using a ‘layered least squares’ (LLS) step\, a
nd showed that O(n^3.5log(chi(A))) iterations suffice to solve (LP) exactl
y\, where chi(A) is a condition measure controlling the size of solutions
to linear systems related to A.

Monteiro and Tsuchiya\, noting tha t the central path is invariant under rescalings of the columns of A and c \, asked whether there exists an LP algorithm depending instead on the mea sure chi*(A)\, defined as the minimum chi(AD) value achievable by a column rescaling AD of A\, and gave strong evidence that this should be the case . We resolve this open question affirmatively.

\nOur first main cont ribution is an O(m^2n^2+n^3) time algorithm which works on the linear matr oid of A to compute a nearly optimal diagonal rescaling D satisfying chi(A D)≤n(chi*(A))3. This algorithm also allows us to approximate the value of chi(A) up to a factor n(chi*(A))2. As our second main contribution\, we de velop a scaling invariant LLS algorithm\, together with a refined potentia l function based analysis for LLS algorithms in general. With this analysi s\, we derive an improved O(n^{2.5}logn log(chi*(A))) iteration bound for optimally solving (LP) using our algorithm. The same argument also yields a factor n/logn improvement on the iteration complexity bound of the origi nal Vavasis-Ye algorithm.

\n>

DTSTART;TZID=America/New_York:20200227T133000 DTEND;TZID=America/New_York:20200227T143000 SEQUENCE:0 SUMMARY:AMS Seminar: Daniel Dadush (Centrum Wiskunde & Informatica (CWI)\, Netherlands) @ Whitehead 304 URL:https://engineering.jhu.edu/ams/events/ams-seminar-tba-whitehead-304-8/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-14934@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title: **Air Pollution Accountability Studies
: Lesson Learned and Future Opportunities

**Abstract: Policy makers seek tools to quantify the net health benefits of improve
d air quality or of proposed air quality regulations. The most commonly ap
plied approach is risk assessment\, that is\, estimating health benefits f
rom expected or observed air quality changes by extrapolating exposure res
ponse functions from existing epidemiologic studies. Accountability studie
s attempt to validate these assessments based on empirical evidence of the
effects on air pollution and health of regulatory actions\, interventions
\, or “natural” experiments. Accountability studies are appealing in that
they are the closest epidemiologic equivalent to controlled experimental s
tudies\, and thus may provide evidence for causal relationships. Neverthel
ess\, accountability studies must disentangle policy-related changes in ai
r pollution and health from other time-varying factors influencing air pol
lution and/or health. We will examine the range of study designs used in a
ccountability studies and the challenges faced in these studies.**

**<
strong>Bio: **Dr. Douglas W. Dockery is the John L. Loeb and France
s Lehman Research Professor of Environmental Epidemiology in the Departmen
ts of Environmental Health and of Epidemiology at the Harvard TH Chan Scho
ol of Public Health. He was Chair of the Department of Environmental Healt
h (2005-2016) and Director of the Harvard-National Institute of Environmen
tal Health Sciences (NIEHS) Center for Environmental Health Sciences (2008
-2019). He received a B.S. in physics from the University of Maryland\, an
M.S. in meteorology from the Massachusetts Institute of Technology\, and
a ScD in environmental health from the Harvard School of Public Health. Dr
. Dockery has been studying air pollution exposures and their health effec
ts for more than four decades. He served as Principal Investigator of the
Harvard Six Cities Study of the Respiratory Health Effects of Respirable P
articles and Sulfur Oxides. His recent work includes assessment of the hea
lth benefits of air pollution controls. Dr. Dockery has published over two
hundred peer-reviewed articles. His 1993 New England Journal of Medicine
paper on air pollution and mortality in the Harvard Six Cities study is th
e single most cited air pollution paper. In 1998\, the International Socie
ty of Environmental Epidemiology honored with the first John Goldsmith Awa
rd from for Outstanding Contributions to the field.

**Title: **Emergent Behavior in Collective Dyna
mics

**Abstract:** A fascinating aspect of collective
dynamics is the self-organization of small-scales

and their emerg ence as higher-order patterns — clusters\, flocks\, tissues\, parties.

\nThe emergence of different patterns can be described in terms of few fundamental “rules of interactions”.

\nI will discuss recent results of the large-time\, large-crowd dynamics\, driven by anticipation that te nd to align the crowd\,

\nwhile other pairwise interactions keep the crowd together and prevent over-crowding.

\nIn particular\, I addre ss the question how short-range interactions lead to the emergence of long -range patterns\,

\ncomparing different rules of interactions based on geometric vs. topological neighborhoods.

DTSTART;TZID=America/New_York:20200312T133000 DTEND;TZID=America/New_York:20200312T143000 SEQUENCE:0 SUMMARY:Cancelled- AMS Seminar: Eitan Tadmore (University of Maryland) @ Wh itehead 304 URL:https://engineering.jhu.edu/ams/events/ams-seminar-eitan-tadmore-univer sity-of-maryland-whitehead-304/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-14979@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title:** TBA

\n

**Abstra
ct:** TBA

**Title:** A Geometric Understanding of Deep Le
arning

**Abstract:** This work introduces an optimal
transportation (OT) view of generative adversarial networks (GANs). Natura
l datasets have intrinsic patterns\, which can be summarized as the manifo
ld distribution principle: the distribution of a class of data is close to
a low-dimensional manifold. GANs mainly accomplish two tasks: manifold le
arning and probability distribution transformation. The latter can be carr
ied out using the classical OT method. From the OT perspective\, the gener
ator computes the OT map\, while the discriminator computes the Wasserstei
n distance between the generated data distribution and the real data distr
ibution\; both can be reduced to a convex geometric optimization process.
Furthermore\, OT theory discovers the intrinsic collaborative—instead of c
ompetitive—relation between the generator and the discriminator\, and the
fundamental reason for mode collapse. We also propose a novel generative m
odel\, which uses an autoencoder (AE) for manifold learning and OT map for
probability distribution transformation. This AE–OT model improves the th
eoretical rigor and transparency\, as well as the computational stability
and efficiency\; in particular\, it eliminates the mode collapse. The expe
rimental results validate our hypothesis\, and demonstrate the advantages
of our proposed model.

**Title: **Complexity of cutting plane and bran
ch-and-bound algorithms

**Abstract: **We present some
results on the theoretical complexity of branch-and-bound (BB) and cuttin
g plane (CP) algorithms for integer programming (linear and nonlinear). We
will first give an exposition of connections between these ideas and prob
lems in mathematical logic and proof theory. We will then present recent r
esults that shed some new light on the efficiency of these two methods\, w
ith quantitative upper and lower bounds on the power of these methods.

The second part of the talk will be based on work done in collaborati on with Hongyi Jiang\, an AMS Ph.D. student\, and Marco Di Summa and Miche le Conforti at the University of Padova.

\n\n

Topic: AMS Week
ly Seminar

\nTime: Apr 9\, 2020 01:30 PM Eastern Time (US and Canada)

Join Zoom Meeting

\nhttps://wse.zoom.us/j/907100613

Meeting ID: 907 100 613

\nOne tap mobile

\n+16465588656\,\,907100613# US (New York)

\n+13126266799\,\,907100613# US (Chicago)

Dial by your location

\n+1 646 558 8656 US (New York)

\n+1 312 626 6799 US (Chicago)

\n+1 669 900 6833 US (San Jose)

\n+1 253 215 8782 US

\n+1 301
715 8592 US

\n+1 346 248 7799 US (Houston)

\nMeeting ID: 907 100
613

\nFind your local number: https://wse.zoom.us/u/acgRGEZiLc

Join by SIP

\n907100613@zoomcrc.com

DTSTART;TZID=America/New_York:20200409T133000 DTEND;TZID=America/New_York:20200409T143000 SEQUENCE:0 SUMMARY:AMS Weekly Seminar w/ Amitabh Basu on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-tba-whitehead-304-20 / X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-14992@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:

**Title: **Decoding When and How to Treat Patie
nts using a Bayesian Probabilistic Reinforcement Learning Approach

**Abstract: **Patients that undergo kidney transplantation a
re at risk for a number of complications and graft rejection after surgery
\, which could lead to death. In order to prevent graft rejection\, immuno
-suppressive therapy such as tacrolimus is administered to patients post-s
urgery. The patients are monitored over time with repeated follow-up recor
ds (e.g.\, tacrolimus blood levels\, creatinine levels\, BMI) after transp
lantation and the dosage levels of the immunosuppresive drugs can be adjus
ted by the clinician. Based on patients’ baseline information and the foll
owup data\, we develop a Bayesian probabilistic reinforcement learning fr
amework to construct an optimal longitudinal treatment strategy for each i
ndividual by combining a longitudinal model for patients’ creatinine level
s\, a survival model with the endpoint being patient death or graft failur
e\, and a marked point process for clinical decisions (how often the patie
nt is instructed to followup\, and drug dosage adjustments). Our method s
hows promising performance on a real kidney transplantation dataset.

\n

*************************************************************

\nTopic: AMS Weekly Seminar

\nTime: Apr 16\, 2020 01:30 PM Eas
tern Time (US and Canada)

Join Zoom Meeting

\nhttps://wse.zoom.us/j/907100613

Mee ting ID: 907 100 613

\nOne tap mobile

\n+16465588656\,\,9071006
13# US (New York)

\n+13126266799\,\,907100613# US (Chicago)

D
ial by your location

\n+1 646 558 8656 US (New York)

\n+1 312 62
6 6799 US (Chicago)

\n+1 669 900 6833 US (San Jose)

\n+1 253 215
8782 US

\n+1 301 715 8592 US

\n+1 346 248 7799 US (Houston)

\nMeeting ID: 907 100 613

\nFind your local number: https://wse.zoo
m.us/u/acgRGEZiLc

Join by SIP

\n907100613@zoomcrc.com

DTSTART;TZID=America/New_York:20200416T133000 DTEND;TZID=America/New_York:20200416T143000 SEQUENCE:0 SUMMARY:AMS Weekly Seminar w/ Yanxun Xu on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-tba-whitehead-304-21 / X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-14999@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:

**Title:** Inference for multiple heterogeneou
s networks with a common invariant subspace

**Abstract: The development of models for multiple heterogeneous network data is
of critical importance both in statistical network theory and across multi
ple application domains. Although single-graph inference is well-studied\,
multiple graph inference is largely unexplored\, in part because of the c
hallenges inherent in appropriately modeling graph differences and yet ret
aining sufficient model simplicity to render estimation feasible. The comm
on subspace independent-edge (COSIE) multiple random graph model addresses
this gap\, by describing a heterogeneous collection of networks with a sh
ared latent structure on the vertices but potentially different connectivi
ty patterns for each graph. The COSIE model is both flexible to account fo
r important graph differences and tractable to allow for accurate spectral
inference. In both simulated and real data\, the model can be deployed fo
r a number of subsequent network inference tasks\, including dimensionalit
y reduction\, classification\, hypothesis testing\, and community detectio
n.**

**Topic: AMS Weekly Seminar**

\nTopic: AMS Weekly Seminar - 4/23

\nTime: Apr 23\, 2020 01:30 PM Eastern Time (US and Canada)\n

Join Zoom Meeting

\nhttps://wse.zoom.us/j/93141026679

\nMeeting ID: 931 4102 6679

\nOne tap mobile

\n+16465588656\,\,93141026679# US (New York)

\n+13126266799\,\,93141026679# US (Chicago)

\n\n

Dial by your location

\n+1 646 558 8656 US (New York)

\n+1 312 626 6799 US (Chicago)

\n+1 301 715 8592 US

\n+1 346 248 7799 US (H ouston)

\n+1 669 900 6833 US (San Jose)

\n+1 253 215 8782 US\n

Meeting ID: 931 4102 6679

\nFind your local number: https://wse.zoom.us/u/acgRGEZiLc

\nJoin by SIP

\n\nJoin by H.323

\n162.255.37.11 (US West)< /p>\n

162.255.36.11 (US East)

\n221.122.88.195 (China)

\n115 .114.131.7 (India Mumbai)

\n115.114.115.7 (India Hyderabad)

\n213.19.144.110 (EMEA)

\n103.122.166.55 (Australia)

\n209.9.211 .110 (Hong Kong

\nChina)

\n64.211.144.160 (Brazil)

\n69. 174.57.160 (Canada)

\n207.226.132.110 (Japan)

\nMeeting ID: 93 1 4102 6679

DTSTART;TZID=America/New_York:20200423T133000 DTEND;TZID=America/New_York:20200423T143000 SEQUENCE:0 SUMMARY:AMS Weekly Seminar w/ Jesus Arroyo on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-tba-whitehead-304-22 / X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-27529@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title: ** Circadian Event Streams: Initial Mo
dels for Prediction and Control

**Abstract:** Our da
ily habits can have long-term health consequences. We are evaluating a si
mple idea for health: eating within a shorter interval of time (for exampl
e 6-8 hours) each day rather than the often found 10-14 hours of eating.
To evaluate this idea\, with funding from the AHA\, we have developed an A
pp (Daily24) and have collected daily times for meals and sleep from more
than 500 individuals over six months of time. This event data can be inte
rpreted as repeated samples from each individuals circadian day\, as an ev
ent stream. If multiple days (like the movie Groundhog Day) are essential
ly similar realizations of each person’s habits\, then we can build up a c
ircadian event model to summarize the data for each individual. We are cu
rrently evaluating a simple mixed component model\, Gaussian Process model
s and state space models. With these models in place we can imagine build
ing a dynamic Hawkes graph to help evaluate circadian-based decisions for
optimizing health by providing predictive feedback on choices for the dail
y habits involved with eating and sleeping..

********************* ************************************************************************** *******

\nTopic: AMS Weekly Seminar w/ Tom Woolf

\nTime: Apr 30
\, 2020 01:30 PM Eastern Time (US and Canada)

Join Zoom Meeting

\nhttps://wse.zoom.us/j/948
39739367

Meeting ID: 948 3973 9367

\nOne tap mobile

\n+16465588656\,\,94839739367# US (New York)

\n+13126266799\,\,948397
39367# US (Chicago)

Dial by your location

\n+1 646 558 8656 U
S (New York)

\n+1 312 626 6799 US (Chicago)

\n+1 301 715 8592 US

\n+1 346 248 7799 US (Houston)

\n+1 669 900 6833 US (San Jose)<
br />\n+1 253 215 8782 US

\nMeeting ID: 948 3973 9367

\nFind you
r local number: https://wse.zoom.us/u/acgRGEZiLc

Join by SIP

\n94839739367@zoomcrc.com

Join by H.323

\n162.255.37.11 (US W
est)

\n162.255.36.11 (US East)

\n221.122.88.195 (China)

\n1
15.114.131.7 (India Mumbai)

\n115.114.115.7 (India Hyderabad)

\n
213.19.144.110 (EMEA)

\n103.122.166.55 (Australia)

\n209.9.211.1
10 (Hong Kong

\nChina)

\n64.211.144.160 (Brazil)

\n69.174.5
7.160 (Canada)

\n207.226.132.110 (Japan)

\nMeeting ID: 948 3973
9367

It is a pleasure for me to kick off this semester’s AMS semi
nar series this coming **Thursday\, Sept 3 at 1:30pm**. This
first seminar is going to be a meet-and-greet\, with a few announcements a
bout how the seminar is going to run this semester in a fully online manne
r. We do intend to have speakers virtually visit Hopkins and even meet and
discuss with people. It will just be in a different format. More details
on Thursday!

The following is the passcode protected link for you to access the Zoom meeting. This is recurring meeting\, so the same link s hould be used every Thursday this Fall. For students\, the following infor mation should also be available from the Blackboard page for the Departmen t seminar EN.553.801.01.F20.

\nhttps://wse.zoom.us/j/98200438 645?pwd=d3M3WEljc0sxd3BRQldUU3dudzhvdz09

\n\n

In case it does not work\, please use the following information:

\nMeeting ID: 982 0043 8645

\nPasscode: 374212

\nTo avoid instances of zoom- bombing\, please do not share the link above with anyone else.

\n\n

**Important: **We still have not been able to fill our
**Sept 10 slot** for the seminar. So **if one of you c
an save the day and give a cool scientific talk **to kick us off ne
xt week\, that would be awesome. Please email me if you are interested and
available.

See you all this Thursday at 1:30pm!

\nDTSTART;TZID=America/New_York:20200903T133000 DTEND;TZID=America/New_York:20200903T143000 SEQUENCE:0 SUMMARY:AMS Weekly Zoom Seminar- Meet & Greet URL:https://engineering.jhu.edu/ams/events/ams-weekly-zoom-seminar-meet-gre et/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28002@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:

**Title**: Numerical tolerance for spectral dec
ompositions of random matrices

\n

**Abstract**
: The computation of parametric estimates often involves iterative numeric
al approximations\, which introduce numerical error. But when these estima
tes depend on random observations\, they necessarily involve statistical e
rror as well. Thus the common approach of minimizing numerical error witho
ut accounting for inherent statistical error can be both costly and wastef
ul\, since it results in no improvement to the estimator’s accuracy. We qu
antify this tradeoff between numerical and statistical error in a problem
of estimating the eigendecomposition for the mean of a random matrix from
its observed value\, and show that one can save significant computation by
terminating the iterative procedure early\, with no loss of accuracy. We
demonstrate this in a setting of estimating the latent positions of a rand
om network from the observed adjacency matrix\, on real and simulated data
.

\n

Your cloud recording is now available.

\nTopic:
AMS Department Seminar (Fall 2020)

\nDate: Sep 10\, 2020 12:18 PM Eas
tern Time (US and Canada)

For host only\, click here to view your
recording (Viewers cannot access this page):

\nhttps://w
se.zoom.us/recording/detail?meeting_id=o3QrttwgRpWP7tUxvIeD0g%3D%3D

Share recording with viewers:

\nhttps://wse.zoom.us/rec/share/4AKeQRT7O46d3cCsr-82-YqVzqf
i58sHJ42n-zFBIQscU7jFBSIzNelTMzVA7GXP.IR-GocHrS2lpCmpH Passcode: L+58i
B^b

**Title**: Looking Forward to Backward-Looking
Rates: A Modeling Framework for Term Rates Replacing LIBOR

\n

**Abstract**: LIBOR and other similar IBOR rates represent
the cost of short-term funding among large global banks\, and are the ref
erence rates in millions of financial contracts with a total market exposu
re worldwide of 400 trillion dollars. Lack of liquidity in the unsecured s
hort-term lending market\, as well as evidence of LIBOR manipulation durin
g the 2007-09 credit crisis\, led regulators to identify new rate benchmar
ks. In this talk\, we introduce and model the new new interest-rate benchm
arks and their compounded setting-in-arrears term rates\, which will be re
placing IBORs globally. We show that the classic interest-rate modeling fr
amework can be naturally extended to describe the evolution of both the fo
rward-looking (IBOR-like) and backward-looking (setting-in-arrears) term r
ates using the same stochastic process. We then introduce an extension of
the LIBOR Market Model to backward-looking rates. Applications will be pre
sented and numerical examples showcased.

\n

Your cloud reco rding is now available.

\nTopic: AMS Department Seminar (Fall 2020)< br />\nDate: Sep 17\, 2020 01:18 PM Eastern Time (US and Canada)

\nF
or host only\, click here to view your recording (Viewers cannot access th
is page):

\nhttps://wse.zoom.us/recording/detail?m
eeting_id=VX%2FqA9N%2FQ%2BynIoBw1R9Mzg%3D%3D

Share recording w
ith viewers:

\nhttps:
//wse.zoom.us/rec/share/CBAf80Hb_1ZlYLpz8DoKhdOwx7k9F1zOsmr4EUdXV9LTgmF5TN
ou-ugp9RkERWlP.bTMc0SwGWnbz4dqY Passcode: uL5&+@!1

**Title:** Ingredients matter: Quick and easy r
ecipes for estimating clusters\, manifolds\, and epidemics

**Abstract:** Data science resembles the culinary arts in the sense
that better ingredients allow for better results. We consider three instan
ces of this phenomenon. First\, we estimate clusters in graphs\, and we fi
nd that more signal allows for faster estimation. Here\, “signal” refers t
o having more edges within planted communities than across communities. Ne
xt\, in the context of manifolds\, we find that an informative prior allow
s for estimates of lower error. In particular\, we apply the prior that th
e unknown manifold enjoys a large\, unknown symmetry group. Finally\, we c
onsider the problem of estimating parameters in epidemiological models\, w
here we find that a certain diversity of data allows one to design estimat
ion algorithms with provable guarantees. In this case\, data diversity ref
ers to certain combinatorial features of the social network. Joint work wi
th Jameson Cahill\, Charles Clum\, Hans Parshall\, and Kaiying Xie.

Your cloud recording is now available.

\nTopic: AMS Depa
rtment Seminar (Fall 2020)

\nDate: Sep 24\, 2020 12:59 PM Eastern Tim
e (US and Canada)

For host only\, click here to view your recordin
g (Viewers cannot access this page):

\nhttps://wse.zoo
m.us/recording/detail?meeting_id=D3Hbv%2Fe5QXKcE0FUgQFdVg%3D%3D

\nhttps://wse.zoom.us/rec/share/fChPLSraWeF5AhXKbY0jkOOfv0zAhnX 4d6qWeWVa9_Goyup0aLcKi0VETt7T2Wan.xDWyUYFDujlhPvqt Passcode: 79W*iV@G< /p> DTSTART;TZID=America/New_York:20200924T133000 DTEND;TZID=America/New_York:20200924T143000 SEQUENCE:0 SUMMARY:AMS Seminar w/ Dustin Mixon (Ohio State University) on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-w-dustin-mixon-ohio- state-university-on-zoom/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28038@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:

**Title**: Learning with entropy-regularized op
timal transport

**Abstract**: Entropy-regularized OT
(EOT) was first introduced by Cuturi in 2013 as a solution to the computat
ional burden of OT for machine learning problems. In this talk\, after stu
dying the properties of EOT\, we will introduce a new family of losses bet
ween probability measures called Sinkhorn Divergences. Based on EOT\, this
family of losses actually interpolates between OT (no regularization) and
MMD (infinite regularization). We will illustrate these theoretical claim
s on a set of learning problems formulated as minimizations over the space
of measures.

\n

Your cloud recording is now available.

\nTopic: AMS Department Seminar (Fall 2020)

\nDate: Oct 1\, 2020 0
1:21 PM Eastern Time (US and Canada)

For host only\, click here to
view your recording (Viewers cannot access this page):

\nhttps://wse.zoom.us/recording/detail?meeting_id=zXLatYK3QFieUi0kc9N%2B
RA%3D%3D

Share recording with viewers:

\nhttps://wse.zoom.us/rec/share/cuYXVU99jAda
Luq4FfIew8x7dxjZ40hORkqQyQpfPCAB_B69q1XeDJmLFw5yuZrb.QIj2wn6azpc4V96E
Passcode: *$xMJcX6

**Title: **Subgraph isomorphism via partial dif
ferentiation

**Abstract: **In this talk I will discus
s a recent approach to the algorithmic problem of* subgraph isomorphism
*: given a host graph G and target graph H\, decide whether G contains
a subgraph isomorphic to H. For simplicity\, I will illustrate the approa
ch in the case when H is a path. I will describe an algorithm whose runtim
e comes close to that of the state of the art\, while using a new approach
based on identifying polynomials with prescribed combinatorial supports (
i.e.\, monomials appearing with nonzero coefficient)\, and whose partial d
erivatives (of all orders) span a vector space of small dimension. Connect
ions to previous approaches and avenues for further improvement will also
be discussed.

Part of this talk is based on joint work with Cornel ius Brand.

\n\n

Here is the link and the meeting info:

\n< p>https://wse.zoom.us/j/98200438645?pwd=d3M3WEljc0sxd3BRQldUU3dudzh vdz09\nMeeting ID: 982 0043 8645

\nPasscode: 374212

\nDTSTART;TZID=America/New_York:20201008T133000 DTEND;TZID=America/New_York:20201008T143000 SEQUENCE:0 SUMMARY:AMS Seminar w/ Kevin Pratt (Carnegie Mellon University) on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-w-kevin-pratt-carneg ie-mellon-school-on-zoom/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28053@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:

**Title:** A Geometric Understanding of Deep Le
arning

**Abstract:** This work introduces an optimal
transportation (OT) view of generative adversarial networks (GANs). Natura
l datasets have intrinsic patterns\, which can be summarized as the manifo
ld distribution principle: the distribution of a class of data is close to
a low-dimensional manifold. GANs mainly accomplish two tasks: manifold le
arning and probability distribution transformation. The latter can be carr
ied out using the classical OT method. From the OT perspective\, the gener
ator computes the OT map\, while the discriminator computes the Wasserstei
n distance between the generated data distribution and the real data distr
ibution\; both can be reduced to a convex geometric optimization process.
Furthermore\, OT theory discovers the intrinsic collaborative—instead of c
ompetitive—relation between the generator and the discriminator\, and the
fundamental reason for mode collapse. We also propose a novel generative m
odel\, which uses an autoencoder (AE) for manifold learning and OT map for
probability distribution transformation. This AE–OT model improves the th
eoretical rigor and transparency\, as well as the computational stability
and efficiency\; in particular\, it eliminates the mode collapse. The expe
rimental results validate our hypothesis\, and demonstrate the advantages
of our proposed model.

\n

Meeting Recording:

\n\nAccess Passcode: Fc1=nKmE

DTSTART;TZID=America/New_York:20201015T133000 DTEND;TZID=America/New_York:20201015T143000 SEQUENCE:0 SUMMARY:AMS Seminar w/ David Gu (Stony Brook University) on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-w-david-gu-stony-bro ok-on-zoom/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28057@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title: **The Importance of Being Correlated:
Implications of Dependence in Joint Spectral Inference across Multiple Net
works

**Abstract:** Spectral inference on multiple ne
tworks is a rapidly-developing subfield of graph statistics. Recent work h
as demonstrated that joint\, or simultaneous\, spectral embedding of multi
ple independent network realizations can deliver more accurate estimation
than individual spectral decompositions of those same networks. Little att
ention has been paid\, however\, to the network correlation that such join
t embedding procedures necessarily induce. In this paper\, we present a de
tailed analysis of induced correlation in a {\\em generalized omnibus} emb
edding for multiple networks. We show that our embedding procedure is flex
ible and robust\, and\, moreover\, we prove a central limit theorem for th
is embedding and explicitly compute the limiting covariance. We examine ho
w this covariance can impact inference in a network time series\, and we c
onstruct an appropriately calibrated omnibus embedding that can detect cha
nges in real biological networks that previous embedding procedures could
not discern. Our analysis confirms that the effect of induced correlation
can be both subtle and transformative\, with import in theory and practice
.

\n

Your cloud recording is now available.

\nTopic:
AMS Department Seminar (Fall 2020)

\nDate: Oct 29\, 2020 01:18 PM Eas
tern Time (US and Canada)

For host only\, click here to view your
recording (Viewers cannot access this page):

\nhttps:/
/wse.zoom.us/recording/detail?meeting_id=%2FQgtuDojRnaeVqoi0IWcuw%3D%3D

Share recording with viewers:

\nhttps://wse.zoom.us/rec/share/1fETcswYJGsGY6HgXmvs4Xd
1EaAI1ThzZceI3AhmxD6c1g-0dkyxc1QLJ5BJFUd4.bXWPzF3wH4zpoKm6 Passcode: ?
a9s6%xR

**Title:** Trainability and accuracy of artific
ial neural networks

**Abstract: **The methods and mod
els of machine learning (ML) are rapidly becoming de facto tools for the a
nalysis and interpretation of large data sets. Complex classification task
s such as speech and image recognition\, automatic translation\, decision
making\, etc. that were out of reach a decade ago are now routinely perfor
med by computers with a high degree of reliability using (deep) neural net
works. These performances suggest that DNNs may approximate high-dimension
al functions with controllably small errors\, potentially outperforming st
andard interpolation methods based e.g. on Galerkin truncation or finite e
lements that have been the workhorses of scientific computing. In support
of this prospect\, in this talk I will present results about the trainabil
ity and accuracy of neural networks\, obtained by mapping the parameters o
f the network to a system of interacting particles relaxing on a potential
determined by the loss function. This mapping can be used to prove a dyna
mical variant of the universal approximation theorem showing that the opti
mal neural network representation can be attained by (stochastic) gradient
descent\, with a approximation error scaling as the inverse of the networ
k size. I will also show how these findings can be used to accelerate the
training of networks and optimize their architecture\, using e.g nonloca
l transport involving birth/death processes in parameter space.

< /p>\n

Meeting recording link:

\nhttps://wse.zoom.us/rec/share/WO_nf9zgmnKfPniZsBSzECdAdNBp5wi
yMP34tsMNAbb1jgVtgqQAV4YtrJjCGPY7.S1xykYLxepdibbZQ

Passcode: M H!7JDN2

DTSTART;TZID=America/New_York:20201105T133000 DTEND;TZID=America/New_York:20201105T143000 SEQUENCE:0 SUMMARY:AMS Seminar w/ Eric Vanden-Eijnden (New York University) on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-w-eric-vanden-eijnde n-nyu-on-zoom/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28101@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title:** Monte-Carlo methods for high-dimensi
onal problems in quantitative finance

**Abstract:** S
tochastic optimal control has been an effective tool for many problems in
quantitative finance and financial economics. Although it provides much ne
eded quantitative modeling for such problems\, until recently it has been
intractable in high-dimensional settings. However\, several recent studies
report impressive numerical results: Cheredito et al. studied the optimal
stopping problem (a problem closely connected to pricing American-type op
tions in quantitative finance finale) providing tight error bounds and an
efficient algorithm in problems in up to 100 dimensions. Buehler et al.\,
on the other hand\, consider the problem of hedging and again report resu
lts for high-dimensional problems that were intractable. These papers use
a Monte Carlo type algorithm combined with deep neural networks proposed b
y E. Han and Jentzen. In this talk I will outline this approach and discu
ss its properties. Numerical results\, while validating the power of the
method in high dimensions\, also show the dependence on the dimension and
the size of the training data. This is joint work with Max Reppen of Bost
on University.

\n

Here is the link and the meeting info:

\nhttps://wse.zoom.us/j/98200438645?pwd=d3M3WEljc0sxd3BRQldUU3d udzhvdz09

\nMeeting ID: 982 0043 8645

\nPasscode**:
374212**

Enjoy.

DTSTART;TZID=America/New_York:20201112T133000 DTEND;TZID=America/New_York:20201112T143000 SEQUENCE:0 SUMMARY:AMS Seminar w/ Mete Soner (Princeton University) on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-w-mete-soner-princet on-on-zoom/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28111@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Goldman Lecture 11-19-2
020(pdf)**

**Title: **Lifting for Simplicity: Conc
ise Descriptions of Convex Sets

**Abstract:**
A commo
n theme in many areas of mathematics is to find a simpler representation o
f an object indirectly by expressing it as the projection of an object in
some higher-dimensional space. In 1991 Yannakakis proved a remarkable con
nection between a lifted representation of a polytope and the nonnegative
rank of a matrix associated with the polytope. In recent years\, this idea
has been generalized to cone lifts of convex sets\, with applications in\
, and tools coming from\, many areas of mathematics and theoretical comput
er science. This talk will survey the central ideas\, results\, and questi
ons in this field.

**Bio:** Rekha Thomas is th
e Walker Family Endowed Professor of Mathematics at

\nthe University
of Washington. She received her Ph.D. in Operations Research from Cornell
University in 1994 followed by postdoctoral work at Yale and Berlin. Her r
esearch interests are in Optimization and Applied Algebraic Geometry.

**Cloud recording is now available.**

Topic: AMS
Department Seminar (Fall 2020)

\nDate: Nov 19\, 2020 01:09 PM Easter
n Time (US and Canada)

**Passcode: +
$0sH0iT **

DTSTART;TZID=America/New_York:20201119T133000 DTEND;TZID=America/New_York:20201119T143000 SEQUENCE:0 SUMMARY:The Goldman Distinguished Lecture Series: Rekha Thomas (University of Washington\, Seattle) on Zoom URL:https://engineering.jhu.edu/ams/events/the-goldman-distinguished-lectur e-series-rekha-thomas-university-of-washington-on-zoom/ X-COST-TYPE:free X-WP-IMAGES-URL:thumbnail\;https://engineering.jhu.edu/ams/wp-content/uploa ds/2020/09/rekhathomas-300x200.jpg\;245\;163\,medium\;https://engineering. jhu.edu/ams/wp-content/uploads/2020/09/rekhathomas-300x200.jpg\;245\;163\, large\;https://engineering.jhu.edu/ams/wp-content/uploads/2020/09/rekhatho mas-300x200.jpg\;245\;163\,full\;https://engineering.jhu.edu/ams/wp-conten t/uploads/2020/09/rekhathomas-300x200.jpg\;245\;163 END:VEVENT BEGIN:VEVENT UID:ai1ec-28110@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:

**Title: **Beyond Mean-field Limits for Large-s
cale Stochastic Systems

**Abstract: **Many large-scal
e stochastic systems that arise as models in a variety of fields including
neuroscience\, epidemiology\, physics\, engineering and computer science\
, can be described in terms of a large collection of “locally” interacting
Markov chains\, where each particle’s transition rates depend only on the
states of neighboring particles with respect to an underlying (possibly r
andom) graph. Since these dynamics are typically not amenable to exact ana
lysis\, a common paradigm is to instead study a more tractable approximati
on that is asymptotically exact as the number of particles goes to infinit
y in order to gain qualitative insight into the system. A frequently used
approximation is the mean-field approximation\, which works provably well
when the interaction graph is sufficiently dense. However\, it performs qu
ite poorly when the interaction graph is sparse\, which is the case in man
y applications. We describe new asymptotically accurate approximations tha
t can be developed in the latter setting\, and show how they perform in va
rious applications. This is joint work with A. Ganguly.

**
Bio:** Kavita Ramanan is the Roland George Dwight Richardson Univer
sity Professor and Associate Chair at the Division of Applied Mathematics\
, Brown University. Her field of research is probability theory\, stochast
ic processes and their applications. She has received several honors in re
cognition of her research\, including a Guggenheim Fellowship\, a Distingu
ished Alumni Award from IIT-Bombay\, and the Newton Award from the Departm
ent of Defense (DoD)\, all in 2020\, a Simons Fellowship in 2018\, an IMS
Medallion in 2015 and the Erlang Prize from the INFORMS Applied Probabilit
y Society in 2006 for “outstanding contributions to applied probability.”
She serves on multiple editorial boards and is an elected fellow of seve
ral societies\, including AAAS\, AMS\, INFORMS\, IMS and SIAM.

Mor
e information about her can be found at her website:

\nhttps://www.brown.edu/academics/applied-mathematics/faculty/kavita-raman
an/home

\n

Your cloud recording is now available.

\n \nPasscode: b#mJ2P+@

DTSTART;TZID=America/New_York:20201203T133000 DTEND;TZID=America/New_York:20201203T143000 SEQUENCE:0 SUMMARY:The Acheson J. Duncan Lecture Series: AMS Seminar: Kavita Ramanan ( Brown University) on Zoom URL:https://engineering.jhu.edu/ams/events/the-acheson-j-duncan-lecture-ser ies-ams-seminar-kavita-ramanan-brown-university-on-zoom/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28780@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title** – Advancing scalable\, provable optim
ization methods in semidefinite & polynomial programs

**Abst
ract **

Optimization is a broad area with ramifications in many disciplines\, including machine learning\, control theory\, signal pr ocessing\, robotics\, computer vision\, power systems\, and quantum inform ation. I will talk about some novel algorithmic and theoretical results in two broad classes of optimization problems. The first class of problems a re semidefinite programs (SDP). I will present the first polynomial time g uarantees for the Burer-Monteiro method\, which is widely used for solving large scale SDPs. I will also discuss some general guarantees on the qual ity of SDP solutions for parameter estimation problems. The second class o f problems I will consider are polynomial systems. I will introduce a nove l technique for solving polynomial systems that\, by taking advantage of g raphical structure\, is able to outperform existing techniques by orders o f magnitude.

\nDTSTART;TZID=America/New_York:20210108T110000 DTEND;TZID=America/New_York:20210108T120000 SEQUENCE:0 SUMMARY:Special Seminar – Faculty Candidate Diego Cifuentes URL:https://engineering.jhu.edu/ams/events/special-seminar-faculty-candidat e-diego-cifuentes/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28686@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:

**Title** – The Landscape of the Proximal Point
Method for Nonconvex-Nonconcave Minimax Optimization

**Abst
ract **

Minimax optimization has become a central tool for modern machine learning with applications in generative adversarial networ ks\, robust training\, reinforcement learning\, etc. These applications ar e often nonconvex-nonconcave\, but the existing theory is unable to identi fy and deal with the fundamental difficulties this poses. In this talk\, w e will overcome these limitations\, describing the convergence landscape o f the classic proximal point method on nonconvex-nonconcave minimax proble ms. Our key theoretical insight lies in identifying a modified objective\, generalizing the Moreau envelope\, that smoothes the original objective a nd convexifies and concavifies it based on the interaction between the min imizing and maximizing variables. When interaction is sufficiently strong\ , we derive global linear convergence guarantees. When interaction is weak \, we derive local linear convergence guarantees under proper initializati on. Between these two settings\, we show undesirable behaviors like diverg ence and cycling can occur.

\n**Bio
**: Benjamin Grimmer is a PhD student
in Operations Research at Cornell University. He received his BS and MS degrees in Computer Science from Illinois
Institute of Technology. His research
focuses on theoretical foundations

Please email Meg Tu lly – mtully4@jhu.edu for more inform ation

DTSTART;TZID=America/New_York:20210111T103000 DTEND;TZID=America/New_York:20210111T113000 SEQUENCE:0 SUMMARY:Special Seminar – Faculty Candidate Ben Grimmer URL:https://engineering.jhu.edu/ams/events/special-seminar-faculty-candidat e-ben-grimmer/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28689@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title** – Decentralized stochastic gradient descent an
d beyond

~~Stochastic gradient descent (SGD) methods have recentl
y found wide applications in large-scale data analysis\, especially in mac
hine learning. These methods are very attractive to process online streami
ng data as they scan through the dataset only once but still generate solu
tions with acceptable accuracy. However\, it is known that classical SGD m
ethods are ineffective in processing streaming data distributed over multi
-agent network systems (e.g.\, sensor and social networks)\, mainly due to
the high communication costs incurred by these methods. In this talk\, we
present a new class of SGD methods\, referred to as stochastic decentrali
zed communication sliding methods\, which can significantly reduce the ~~aforementioned communication costs for decentralized stochastic optimization and machine lear
ning. We show that these methods can skip inter-node communications while
performing SGD iterations. As a result\, they require a substantially smal
ler number of communication rounds than existing decentralized SGD\, while
the total number of required stochastic subgradient computations are comparable to those optimal
bounds achieved by classical centralized SGD type methods. We also develop
new variants of these methods that can achieve graph topology invariant g
radient/sampling complexity when the problem is smooth and samples can be stor
ed locally.

**BIO**: <
span class='NormalTextRun SpellingErrorV2 BCX9 SCXW79695960'>Guanghui (George) Lan is an associ
ate professor in the H. Milton Stewart School of Industrial and Systems En
gineering at Georgia Institute of Technology since January 2016. Dr. Lan w
as on the faculty of the Department of Industrial and Systems Engineering
at the University of Florida from 2009 to 2015\, after earning his Ph.D. d
egree from Georgia Institute of Technology in August 2009. His main resear
ch interests lie in optimization and machine learning. The academic honors
he received include the Mathematical Optimization Society Tucker Prize Fi
nalist (2012)\, INFORMS Junior Faculty Interest Group Paper Competition Fi
rst Place (2012) and the National Science Foundation CAREER Award (2013).
Dr. Lan serves as an associate editor for Mathematical Programming\, SIAM
Journal on Optimization and Computational Optimization and Applications. H
e is also an associate director of the Center for Machine Learning at Geor
gia Tech.

For Zoom information email Meg Tully – mtully4@jhu.edu

DTSTART;TZID=America/New_York:20210113T100000 DTEND;TZID=America/New_York:20210113T110000 SEQUENCE:0 SUMMARY:Special Seminar – Guanghui (George) Lan URL:https://engineering.jhu.edu/ams/events/special-seminar-guanghui-george- lan/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28690@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title** – On
Complexity of Constrained Nonconvex Optimization

**A
bstract**

Deriving complex ity guarantees for nonconvex optimization problems are driven by long stan ding theoretical interests and by their relevance to machine learning and data science. This talk discusses complexity of algorithms for two importa nt types of constrained nonconvex optimization problems: bound-constrained and nonlinear equality constrained optimization. Applications include non negative matrix factorization (NMF) and dictionary learning.

\n< p> \nFor nonconvex optimization wi th bound constraints\, we observe from the past work that pursuit of the s tate-of-art complexity guarantees can compromise the practicality of an al gorithm. Therefore\, we propose two practical projected Newton types of me thods with complexity guarantees matching the best known. The first method is a scaled variant of Bertsekas’ two-metric projection method\, with the best complexity guarantee to find an approximate first-order point. The s econd is a projected Newton-Conjugate Gradient method\, equipped with a co mpetitive complexity guarantee to locate an approximate second-order point with high probability. Preliminary numerical experiments on NMF indicate practicality of the latter algorithm.

\n\n

For nonconvex optimization with nonlinear equality c onstraints\, we analyze complexity of the proximal augmented Lagrangian (A L) framework\, in which a Newton-Conjugate-Gradient scheme is used to find approximate solutions of the subproblems. This scheme has three levels of iterations\, and we obtain bounds on the number of iterations at each lev el.

\n\n

These are jo int works with Stephen J. Wright.

\n\n

For zoom inform ation email Meg Tully – mtully4@jhu.edu< /a>

DTSTART;TZID=America/New_York:20210114T103000 DTEND;TZID=America/New_York:20210114T113000 SEQUENCE:0 SUMMARY:Special Seminar – Faculty Candidate Yue Xie URL:https://engineering.jhu.edu/ams/events/special-seminar-faculty-candidat e-yue-xie/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28784@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title –**~~ Optimal Transport for Invers
e Problems and the Implicit Regularization~~

**A
bstract **

Optimal transport has been one interesting topic of mathema tical analysis since Monge (1781). The problem’s close connections with di fferential geometry and kinetic descriptions were discovered within the pa st century\, and the seminal work of Kantorovich (1942) showed its power t o solve real-world problems. Recently\, we proposed the quadratic Wasserst ein distance from optimal transport theory for inverse problems\, tackling the classical least-squares method’s longstanding difficulties such as no nconvexity and noise sensitivity. The work was soon adopted in the oil ind ustry. As we advance\, we discover that the advantage of changing the data misfit is more general in a broader class of data-fitting problems by exa mining the preconditioning and “implicit” regularization effects of differ ent mathematical metrics as the objective function in optimization\, as th e likelihood function in Bayesian inference\, and as the measure of residu al in numerical solution to PDEs.

DTSTART;TZID=America/New_York:20210119T100000 DTEND;TZID=America/New_York:20210119T110000 SEQUENCE:0 SUMMARY:Special Seminar – Faculty Candidate Yunan Yang URL:https://engineering.jhu.edu/ams/events/special-seminar-faculty-candidat e-yunan-yang/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28692@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title** – Data-driven optimization: test scor
e algorithms and distributionally robust approach

**Abstract
**

The ever-increasing availability of data has motivated novel optimization models that can incorporate uncertainty in problem para meters. Data-driven optimization frameworks integrate sophisticated statis tical estimation parts into optimization frameworks. This often leads to c omputationally challenging formulations that mandate new technologies to a ddress scalability issues. For combinatorial optimization\, algorithms nee d to be adapted based on parameter estimation schemes\, and at the same ti me\, need to provide near-optimal performance guarantees even under uncert ain parameters. For mathematical programming models\, efficient solution m ethods are required to deal with added complexity from applying data-drive n frameworks. In this talk\, we discuss test score algorithms for stochast ic utility maximization and talk about distributionally robust chance-cons trained programming.

\nTest score algorithms are based on carefully designed score metrics to evaluate individual items\, called test scores\, defined as a statistic of observed individual item performance data. Algo rithms based on individual item scores are practical when evaluating diffe rent combinations of items is difficult. We show that a natural greedy alg orithm that selects items solely based on their test scores outputs soluti ons within a constant factor of the optimum for a broad class of utility f unctions. Our algorithms and approximation guarantees assume that test sco res are noisy estimates of certain expected values with respect to margina l distributions of individual item values\, thus making our algorithms pra ctical.For the second part of the talk\, we consider distributionally robu st optimization (DRO) frameworks\, which allow interpolating between tradi tional robust optimization and stochastic optimization\, thereby providing a systematic way of hedging against the ambiguity in underlying probabili ty distributions. In particular\, we apply the DRO framework defined with the Wasserstein distance to chance-constrained programming (CCP)\, an opti mization paradigm that involves constraints that have to be satisfied with high probability. We develop formulations by revealing hidden connections between the Wasserstein DRO framework and its nominal counterpart (the sa mple average approximation)\, and propose integer programming based soluti on methods. Our formulations significantly scale up the problem sizes that can be handled by reducing the solution times from hours to seconds\,comp ared to the existing formulations.

\nThis talk is based on joint wor ks with Nam Ho-Nguyen\, Fatma Kılınç-Karzan\, Simge Küçükyavuz\, MilanVojn ovic\, and Se-young Yun.

\nFor zoom information email Meg Tully – mtully4@jhu.edu

DTSTART;TZID=America/New_York:20210120T090000 DTEND;TZID=America/New_York:20210120T100000 SEQUENCE:0 SUMMARY:Special Seminar – Faculty Candidate Debeen Lee URL:https://engineering.jhu.edu/ams/events/special-seminar-faculty-candidat e-debeen-lee/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28790@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title** – Novel Optimization for Data-Driven Decisio
n Making

**Abstract**

We present several exciting works on data-driven decision making. First\, during the crisis o f COVID-19\, we have seen the importance of clinical trial designs. Improv ing clinical trial designs is important for the wellness of all human bein gs. We first present a novel optimization framework for adaptive trial des ign in the context of personalized medicine. Adaptive enrichment designs i nvolve preplanned rules for modifying enrollment criteria based on accruin g data in a randomized trial. We focus on designs where the overall popula tion is partitioned into two predefined subpopulations\, e.g.\, based on a biomarker or risk score measured at baseline for personalized medicine. T he goal is to learn which populations benefit from an experimental treatme nt. Two critical components of adaptive enrichment designs are the decisio n rule for modifying enrollment and multiple testing procedures. We provid e a general framework for simultaneously optimizing these components for t wo-stage\, adaptive enrichment designs through Bayesian optimization. We m inimize the expected sample size under constraints on power and the family wise Type I error rate. It is computationally infeasible to directly solve this optimization problem due to its nonconvexity and infinite dimensiona lity. The key to our approach is a novel\, discrete representation of this optimization problem as a sparse linear program\, which is large-scale bu t computationally feasible to solve using modern optimization techniques. Applications of our approach produce new\, approximately optimal designs. We then present some other optimal data-driven decision mak ing works on high-dim ensional linear contextual bandit and reinforcement learning problems.

DTSTART;TZID=America/New_York:20210121T103000 DTEND;TZID=America/New_York:20210121T113000 SEQUENCE:0 SUMMARY:Special Seminar – Ethan Fang URL:https://engineering.jhu.edu/ams/events/special-seminar-ethan-fang/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28794@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title** – Polar deconvolution of mixed signal
s

**Abstract **

The signal demixing problem s eeks to separate multiple signals from their superposition. I will describ e a geometric view of the superposition process\, and how the duality of c onvex cones allows us to develop an efficient algorithm for recovering the components with sublinear iteration complexity and linear storage. Under a random measurement model\, this process stably recovers low-complexity a nd incoherent signals with high probability and with optimal sample comple xity. This is joint work with my students and postdocs Zhenan Fan\, Halyun Jeong\, and Babhru Joshi.

\nDTSTART;TZID=America/New_York:20210122T110000 DTEND;TZID=America/New_York:20210122T120000 SEQUENCE:0 SUMMARY:Special Seminar – Michael Friedlander URL:https://engineering.jhu.edu/ams/events/special-seminar-michael-friedlan der/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28897@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:

**TITLE:** Short simplex paths in l
attice polytopes

**ABSTRACT: We design a simplex algorithm for linear programs on latti
ce polytopes that traces “short” simplex paths from any given vertex to an
optimal one. We consider a lattice polytope P contained in [0\, k]^n and
defined via ‘m’ linear inequalities. Our first contribution is a simplex a
lgorithm that reaches an optimal vertex by tracing a path along the edges
of P of length in O(n^4 k log(n k)). The length of this path is independen
t of ‘m’ and is only polynomially far from the worst-case diameter\, which
roughly grows as nk. **

Motivated by the fact th at most known lattice polytopes are defined via 0\,+1\,-1 constraint matri ces\, our second contribution is a more sophisticated simplex algorithm wh ich exploits the largest absolute value of the entries in the constraint m atrix\, denoted by ‘a’. We show that the length of the simplex path genera ted by this algorithm is in O(n^2 k log(n k a)). In particular\, if the pa rameter ‘a’ is bounded by a polynomial in n\, k\, then the length of the s implex path is in O(n^2 k log(n k)). This is a joint work with Alberto Del Pia.

\n\n

The cloud recording is now available.

\nTopic: AMS Department Seminar (Spring 2021)

\nDate: Feb 4\, 2021 0
1:19 PM Eastern Time (US and Canada)

Recording for viewers:

\nhttps://wse.zoom.us/rec/
share/mr4m196sKhWTcUV4TYKMWcT1MUxgUI7KdFoUhRxxrPBqew_OVKX0X7kG_Lee4jKN.kCm
Cor9FOsMzTuL5 Passcode: qB4+9D6q

Enjoy.

DTSTART;TZID=America/New_York:20210204T133000 DTEND;TZID=America/New_York:20210204T143000 SEQUENCE:0 SUMMARY:AMS Seminar w/ Carla Michini (University of Wisconsin-Madison) on Z oom URL:https://engineering.jhu.edu/ams/events/28897-2/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28903@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title:** Quantile-based Iterative Methods for
Corrupted Systems of Linear Equations

**Abstract: **
One of the most ubiquitous problems arising across the sciences is that of
solving large-scale systems of linear equations Ax = b. When it is infeas
ible to solve the system directly by inversion\, scalable and efficient pr
ojection-based iterative methods can be used instead\, such as\, Randomize
d Kaczmarz (RK) algorithm\, or SGD (optimizing ||Ax – b|| in some norm).\n

The main goal of my talk is to present versions of these two algori thms aimed at linear systems with adversarially corrupted vector b\, Quant ileRK and QuantileSGD. While the classical approach for noisy (inconsisten t) systems is to show that the iterations approach the least squares solut ion until a certain convergence horizon that depends on the noise size\, i n order to handle large\, sparse\, potentially adversarial corruptions\, o ne needs to modify the algorithm to avoid corruptions rather than try to t olerate them — and quantiles of the residual provide a natural way to do s o. Our methods work on up to 50% of incoherent corruptions\, and up to 20% of adversarial corruptions (that consistently create an “alternative” sol ution of the system). Theoretically\, under some standard assumptions on t he measurement model\, despite corruptions of any size both methods conver ge to the true solution with exactly the same rate as RK on an uncorrupted system up to an absolute constant. Our theoretical analysis is based on p robabilistic concentration of measure results\, and as an auxiliary random matrix theory result\, we prove a non-trivial uniform bound for the small est singular values of all submatrices of a given matrix. Based on the joi nt work with Jamie Haddock\, Deanna Needell\, and Will Swartworth.

\n\n

The cloud recording is now available.

\nTopic: AMS Depart
ment Seminar (Spring 2021)

\nDate: Feb 11\, 2021 12:49 PM Eastern Tim
e (US and Canada)

Share recording with viewers:

\nhttps://wse.zoom.us/rec/share/Dn4mMX4
n4CMPKWUQfXql6Trt4yurYBxPe37OkFowp0Fcxp3MfNEsnczBxr68roig.90xVqiC1mKI-98wl
Passcode: Jx*#6Lko

Enjoy.

DTSTART;TZID=America/New_York:20210211T133000 DTEND;TZID=America/New_York:20210211T143000 SEQUENCE:0 SUMMARY:AMS Seminar w/ Liza Rebrova (University of California\, Los Angeles ) on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-w-liza-rebrova-unive rsity-of-california-los-angeles-on-zoom/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28905@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title:** Data Integration: Data-Driven Discov
ery from Diverse Data Sources

**Abstract:** Data inte
gration\, or the strategic analysis of multiple sources of data simultaneo
usly\, can often lead to discoveries that may be hidden in individual anal
yses of a single data source. In this talk\, we present several new techn
iques for data integration of mixed\, multi-view data where multiple sets
of features\, possibly each of a different domain\, are measured for the s
ame set of samples. This type of data is common in healthcare\, biomedici
ne\, national security\, multi-senor recordings\, multi-modal imaging\, an
d online advertising\, among others. In this talk\, we specifically highli
ght how mixed graphical models and new feature selection techniques for mi
xed\, multi-view data allow us to explore relationships amongst features f
rom different domains. Next\, we present new frameworks for integrated pr
incipal components analysis and integrated generalized convex clustering t
hat leverage diverse data sources to discover joint patterns amongst the s
amples. We apply these techniques to integrative genomic studies in cance
r and neurodegenerative diseases to make scientific discoveries that would
not be possible from analysis of a single data set.

Here is the n ew link and meeting ID+passcode:

\nhttps://wse.zoom.us/j/9146 7375713?pwd=VjN3ekZTRFZIWS80NnpwZUFRUzRWUT09

\nMeeting ID: 914 6 737 5713

\nPasscode: 272254

\nDTSTART;TZID=America/New_York:20210218T133000 DTEND;TZID=America/New_York:20210218T143000 SEQUENCE:0 SUMMARY:AMS Seminar w/ Genevera Allen (Rice University) on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-w-genevera-allen-ric e-university-on-zoom/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28911@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:

**Title: **S
tatistical Approaches to Studying Air Pollution Mixtures and Health

**Bio:** Dr. Roger D. Peng is a Professor of Biosta
tistics at the Johns Hopkins Bloomberg School of Public Health where his r
esearch focuses on the development of statistical methods for addressing e
nvironmental health problems. He has led some of the largest national stud
ies on the health effects of ambient air pollution in the United States. D
r. Peng is the author of the popular book R Programming for Data Science a
nd 10 other books on data science and statistics. He is also the co-creato
r of the Johns Hopkins Data Science Specialization\, the Simply Statistics
blog where he writes about statistics for the public\, the Not So Standar
d Deviations podcast with Hilary Parker\, and The Effort Report podcast wi
th Elizabeth Matsui. Dr. Peng is a Fellow of the American Statistical Asso
ciation and is the recipient of the Mortimer Spiegelman Award from the Ame
rican Public Health Association\, which honors a statistician who has made
outstanding contributions to public health.

\n

Here is the new link and meeting ID+passcode:

\nhttps://wse.zoom.us/j/91 467375713?pwd=VjN3ekZTRFZIWS80NnpwZUFRUzRWUT09

\nMeeting ID: 914 6737 5713

\nPasscode: 272254

DTSTART;TZID=America/New_York:20210225T133000 DTEND;TZID=America/New_York:20210225T143000 SEQUENCE:0 SUMMARY:The John C. & Susan S.G. Wierman Lecture Series- AMS Seminar w/ Dr. Roger Peng (JHU Biostatistics) on Zoom URL:https://engineering.jhu.edu/ams/events/the-john-c-susan-s-g-wierman-lec ture-series-ams-seminar-w-roger-peng-jhu-biostatistics-on-zoom/ X-COST-TYPE:free X-WP-IMAGES-URL:thumbnail\;https://engineering.jhu.edu/ams/wp-content/uploa ds/2021/02/rpeng_headshot4_sq-150x150.png\;150\;150\;1\,medium\;https://en gineering.jhu.edu/ams/wp-content/uploads/2021/02/rpeng_headshot4_sq-300x30 0.png\;300\;300\;1\,large\;https://engineering.jhu.edu/ams/wp-content/uplo ads/2021/02/rpeng_headshot4_sq.png\;960\;960\; END:VEVENT BEGIN:VEVENT UID:ai1ec-28920@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title: **Denoising as a Building Bl
ock: Form\, function\, and regularization of inverse problems

**Abstract: **Denoising of images has reached impressive l
evels of quality — almost as good as we can ever hope. There are thousands
of papers on this topic\, and their scope is so vast and approaches so di
verse that putting them in some order is useful and challenging. I will sp
eak about why we should still care deeply about this topic\, what we can s
ay about this general class of operators on images\, and what makes them s
o special. Of particular interest is how we can use denoisers as building
blocks for broader image processing tasks\, including as regularizers for
general inverse problems.

**Bio**: Peyman is
a Principal Scientist / Director at Google Research\, where he leads the C
omputational Imaging team. Prior to this\, he was a Professor of Electrica
l Engineering at UC Santa Cruz from 1999-2014. He was Associate Dean for R
esearch at the School of Engineering from 2010-12. From 2012-2014 he was o
n leave at Google-x\, where he helped develop the imaging pipeline for Goo
gle Glass. Most recently\, Peyman’s team at Google developed the digital z
oom pipeline for the Pixel phones\, which includes the multi-frame super-r
esolution (“Super Res Zoom”) pipeline\, and the RAISR upscaling algorithm.
In addition\, the Night Sight mode on Pixel 3 uses our Super Res Zoom te
chnology to merge images (whether you zoom or not) for vivid shots in low
light.

Peyman received his undergraduate education in electrical e ngineering and mathematics from the University of California\, Berkeley\, and the MS and PhD degrees in electrical engineering from the Massachusett s Institute of Technology. He holds 15 patents\, several of which are comm ercially licensed. He founded MotionDSP\, which was acquired by Cubic Inc. (NYSE:CUB).

\nPeyman has been keynote speaker at numerous technical conferences including Picture Coding Symposium (PCS)\, SIAM Imaging Scien ces\, SPIE\, and the International Conference on Multimedia (ICME). Along with his students\, he has won several best paper awards from the IEEE Sig nal Processing Society.

\nHe is a Distinguished Lecturer of the IEEE Signal Processing Society\, and a Fellow of the IEEE “for contributions t o inverse problems and super-resolution in imaging.”

\n\n\n

Here is the new li nk and meeting ID+passcode:

\nhttps://wse.zoom.us/j/914673757 13?pwd=VjN3ekZTRFZIWS80NnpwZUFRUzRWUT09

\nMeeting ID: 914 6737 5 713

\nPasscode: 272254

\n\n

DTSTART;TZID=America/New_York:20210304T133000 DTEND;TZID=America/New_York:20210304T143000 SEQUENCE:0 SUMMARY:AMS Seminar w/ Peyman Milanfar (Google) on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-w-peyman-milanfar-go ogle-on-zoom/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28927@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:

**Title:** Real-time Data Fusion to Guide Disea
se Forecasting Models

**Abstract: **Globalization has
created complex problems that can no longer be adequately understood and
mitigated using traditional data analysis techniques and data sources. As
such\, there is a need for the integration of nontraditional data streams
and approaches such as social media and machine learning to address these
new challenges. In this talk\, I will discuss how our team is applying app
roaches from the weather forecasting community including data collection\,
assimilating heterogeneous data streams into models\, and quantifying unc
ertainty to forecast infectious diseases like COVID-19. In addition\, I w
ill demonstrate that although epidemic forecasting is still in its infancy
\, it’s a growing field with great potential and mathematical modeling wil
l play a key role in making this happen.

\n

Here is the new link and meeting ID+passcode:

\nhttps://wse.zoom.us/j/914673 75713?pwd=VjN3ekZTRFZIWS80NnpwZUFRUzRWUT09

\nMeeting ID: 914 673 7 5713

\nPasscode: 272254

DTSTART;TZID=America/New_York:20210311T133000 DTEND;TZID=America/New_York:20210311T143000 SEQUENCE:0 SUMMARY:AMS Seminar w/ Sara Del Valle (Los Alamos National Labs) on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-w-sara-del-valle-los -alamos-national-labs-on-zoom/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28933@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title:** Memory lower bounds for sampling

**Abstract:** Suppose we would like to maintain a (multi
)subset S of {1\,…\,n} dynamically subject to items being inserted into an
d deleted from S. Then when a user says “sample()”\, we should return a (u
niformly) random element of S\, or an easier task\, return just some (any)
element in S. How much memory is required to accomplish this task? We ans
wer

\nthis question by giving an asymptotically optimal lower bound o
n the memory required.

Joint work with Michael Kapralov\, Jakub Pa
chocki\, Zhengyu Wang\, David

\nP. Woodruff\, and Mobin Yahyazadeh.\n

\n

Here is the recording from the seminar:

\n\nPasscode: T&^!!4=C

DTSTART;TZID=America/New_York:20210318T133000 DTEND;TZID=America/New_York:20210318T143000 SEQUENCE:0 SUMMARY:AMS Seminar w/ Jelani Nelson ( University of California\, Berkeley) on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-w-jelani-nelson-univ ersity-of-california-berkeley-on-zoom/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28939@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title:** On learning kernels for numerical ap
proximation and learning.

**Abstract: **There is a gr
owing interest in solving numerical approximation problems as learning pro
blems. Popular approaches can be divided into (1) Kernel methods\, and (2)
methods based on variants of Artificial Neural Networks. We illustrate th
e importance of using adapted kernels in kernel methods and discuss strate
gies for learning kernels from data. We show how ANN methods can be formu
lated and analyzed as (1) kernel methods with warping kernels learned from
data\, and (2) discretized solvers for a generalization of image registra
tion algorithms in which images are replaced by high dimensional shapes.\n

\n

Here is the recording for the meeting:

\n\nPasscode: ?tW.6T+%

\nDTSTART;TZID=America/New_York:20210325T133000 DTEND;TZID=America/New_York:20210325T143000 SEQUENCE:0 SUMMARY:AMS Seminar w/ Houman Owhadi (Caltech) on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-w-houman-owhadi-calt ech-on-zoom/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28941@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:

**Title:** Nature vs. Nurture in Complex (and N
ot-So-Complex) Systems

**Abstract: **Understanding th
e dynamical behavior of many-particle systems following a deep quench is a
central issue in both statistical mechanics and complex systems theory. O
ne of the basic questions centers on the issue of predictability: given a
system with a random initial state evolving through a well-defined stochas
tic dynamics\, how much of the information contained in the state at futur
e times depends on the initial condition (“nature”) and how much on the dy
namical realization (“nurture”)? We discuss this question and present both
old and new results for both homogeneous and random systems in low and hi
gh dimension.

Starting from next week\, I’ll be taking over the se minar hosting duties from Amitabh\, who is going on paternity leave. We’ll keep the zoom link and all other procedures exactly the same as they are now. Amitabh has created a well-oiled machine!

\n\n

Here is t he recording from the seminar above:

\n\nPassc ode: ^R+Q1=r3

DTSTART;TZID=America/New_York:20210401T133000 DTEND;TZID=America/New_York:20210401T143000 SEQUENCE:0 SUMMARY:AMS Seminar w/ Daniel Stein (New York University) on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-w-daniel-stein-new-y ork-university-on-zoom/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28943@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title: **Phase Analysis of a Family of Reacti
on-Diffusion Equations

\n

**Abstract: **We con
sider a reaction-diffusion equation driven by multiplicative space-time wh
ite noise\, for a large class of reaction terms that include well-known ex
amples such as the Fisher-KPP and Allen-Cahn equations. We prove that\, in
the “intermittent regime”: (1) If the equation is sufficiently noisy\, th
en the resulting stochastic PDE has a unique invariant measure\; and (2) I
f the equation is in a low-noise regime\, then there are infinitely many i
nvariant measures and the collection of all invariant measures is a line s
egment in path space. This gives proof to earlier predictions of Zimmerman
et al (2000)\, discovered first through experiments and computer simulati
ons.

This is joint work with Carl Mueller (University of Rochester ) and Kunwoo Kim (POSTECH).

\n\n

Here is the new link and mee ting ID+passcode:

\nhttps://wse.zoom.us/j/91467375713?pwd=VjN 3ekZTRFZIWS80NnpwZUFRUzRWUT09

\nMeeting ID: 914 6737 5713

\n< p>Passcode: 272254 DTSTART;TZID=America/New_York:20210408T133000 DTEND;TZID=America/New_York:20210408T143000 SEQUENCE:0 SUMMARY:AMS Seminar w/ Davar Khosnevisan (University of Utah) on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-w-davar-khosnevisan- university-of-utah-on-zoom/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28945@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title:** Finding a Compositional Square Root
of Sine

**Abstract:** We consider the following type
of problem: Given a function g : A ! A\, find a

\nfunction f such tha
t g = f f . We are especially interested in the case sin : R ! R\, but

\nconsider the problem more broadly with results for other functions g
defined on other sets

\nA. This is joint work with JHU undergraduate
Tongtong Chen. And\, despite appearances

\nto the contrary\, this is
a graph theory talk.

\n

Here is the new link and meeting ID +passcode:

\nhttps://wse.zoom.us/j/91467375713?pwd=VjN3ekZTRF ZIWS80NnpwZUFRUzRWUT09

\nMeeting ID: 914 6737 5713

\nPassc ode: 272254

DTSTART;TZID=America/New_York:20210415T133000 DTEND;TZID=America/New_York:20210415T143000 SEQUENCE:0 SUMMARY:AMS Seminar w/ Ed Scheinerman (JHU-AMS) on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-w-ed-scheinerman-jhu -ams-on-zoom/ X-COST-TYPE:free END:VEVENT BEGIN:VEVENT UID:ai1ec-28948@engineering.jhu.edu/ams DTSTAMP:20210419T182835Z CATEGORIES: CONTACT: DESCRIPTION:**Title:** TBA

\n

**Abstra
ct:** TBA

\n

Here is the new link and meeting ID+pas scode:

\nhttps://wse.zoom.us/j/91467375713?pwd=VjN3ekZTRFZIWS 80NnpwZUFRUzRWUT09

\nMeeting ID: 914 6737 5713

\nPasscode: 272254

DTSTART;TZID=America/New_York:20210429T133000 DTEND;TZID=America/New_York:20210429T143000 SEQUENCE:0 SUMMARY:AMS Seminar w/ Jim Gatheral (Baruch College) on Zoom URL:https://engineering.jhu.edu/ams/events/ams-seminar-w-jim-gatheral-baruc h-college-on-zoom/ X-COST-TYPE:free END:VEVENT END:VCALENDAR