Calendar

Nov
29
Thu
Distinguished Lecture Series: Sameer Sonkusale, Tufts University @ Hackerman 320
Nov 29 @ 12:00 pm – 1:00 pm
Dec
11
Tue
Undergraduate Assembly
Dec 11 @ 5:00 pm – 7:30 pm

All ECE undergraduate students should make every effort to attend – we will be taking attendance. In the meantime, you can submit questions for Professor Etienne-Cummings to answer during the assembly.

 

Jan
31
Thu
Distinguished Lecture Series: Andreas Tolias, Baylor College of Medicine @ Clark 110
Jan 31 @ 3:00 pm

Abstract: Despite major advances in artificial intelligence through deep learning methods, computer algorithms remain vastly inferior to mammalian brains, and lack a fundamental feature of animal intelligence: they generalize poorly outside the domain of the data they have been trained on. This results in brittleness (e.g. adversarial attacks) and poor performance in transfer learning, few-shot learning, casual reasoning and scene understanding, as well as difficulty with lifelong and unsupervised learning – all important hallmarks of human intelligence. We conjecture that this gap is caused by the fact that current deep learning architectures are severely under-constrained, lacking key model biases found in the brain that are instantiated by the multitude of cell types, pervasive feedback, innately structured connectivity, specific non-linearities, and local learning rules. There is ample behavioral evidence that the brain performs approximate Bayesian inference under a generative model of the world (also known as inverse graphics or analysis by synthesis), so the brain must have evolved a strong and useful model bias that allows it to efficiently learn such a generative model. Therefore, our goal is to learn the brain’s model bias in order to engineer less artificial, and more intelligent, neural networks. Experimental neuroscience now has technologies that enable us to analyze how brain circuits work in great detail and with impressive breadth. Using tour-de-force experimental methods we have been collecting an unprecedented amount of neural responses (e.g. more than 1.5 million neuron-hours) from the visual cortex, and developed computational models that we use to extract principles of functional organization of the brain and learn the brain’s model biases.

 

Biography: Dr. Andreas Tolias’ research goal is to decipher brain’s mechanisms of intelligence. He studies how networks of neurons are structurally and functionally organized to process information. Research in his lab combines computational and machine learning approaches to electrophysiological (whole-cell and multi-electrode extracellular), multi-photon imaging, molecular and behavioral methods. He got his Ph.D. from MIT in Computational and Systems Neuroscience. The current focus of research in his lab is to reverse engineer neocortical intelligence. To this end his lab is deciphering the structure of microcircuits in visual cortex (define cell types and connectivity), elucidate the computations they perform and apply these principles to develop novel machine learning algorithms. He has trained numerous graduate students and postdoctoral fellows and enjoys mentoring immensely.

Feb
7
Thu
Seminar: Ilya Shipster @ Shaffer Hall 2
Feb 7 @ 3:00 pm – 4:00 pm
Feb
14
Thu
Valentines Day Celebration @ ECE Student Lounge
Feb 14 all-day

Valentine Day card making and treats available all day in the ECE Student Lounge!

Distinguished Lecture Series: Victor Klimov, Los Alamos National Laboratory (The Minkowski Lecture) @ Hodson Hall 210
Feb 14 @ 3:00 pm – 4:00 pm

Abstract: Chemically synthesized quantum dots (QDs) can potentially enable new classes of
highly flexible, spectrally tunable lasers processible from solutions [1,2]. Despite a considerable progress over the past years, colloidal-QD lasing, however, is still at the laboratory stage and an important challenge – realization of lasing with electrical injection – is still unresolved. A major complication, which hinders the progress in this field, is fast nonradiative Auger recombination of gain-active multicarrier species such as trions (charged excitons) and biexcitons [3,4]. Recently, we explored several approaches for mitigating the problem of Auger decay by taking advantage of a new generation of core/multi-shell QDs with a radially graded composition that allow for considerable (nearly complete) suppression of Auger recombination by “softening” the electron and hole confinement potentials [5,6]. Using these specially engineered QDs, we have been able to realize optical gain with direct-current electrical pumping [7], which has been a long-standing goal in the field of colloidal nanostructures. Further, we apply these dots to practically demonstrated the viability of a “zero-threshold-optical-gain” concept using not neutral but negatively charged particles wherein the pre-existing electrons block either partially or completely ground-state absorption [8]. Such charged QDs are optical-gain-ready without excitation and, in principle, can exhibit lasing at vanishingly small pump levels. All of these exciting recent developments demonstrate a considerable promise of colloidal nanomaterials for implementing solution-processible optically and electrically pumped laser devices operating across a wide range of wavelengths.

Feb
17
Sun
2019 Engineers Week
Feb 17 – Feb 23 all-day
Feb
21
Thu
Dissertation Defense: Adam Khalifa @ Remsen Hall 101
Feb 21 @ 10:00 am
Proposal: Phillip Wilcox @ Shaffer Hall 2
Feb 21 @ 3:00 pm
Feb
28
Thu
Proposal: Nathan Henry @ Shaffer Hall 2
Feb 28 @ 3:00 pm

Abstract: Frequency combs generated by coherent stimulated emission have revolutionized the precision at which we are able to measure time, frequency, and distance. Directly measuring the frequency of the radiation offers a much higher resolution, as time can be measured with a higher accuracy than distance, and conversion between frequency and wavelength can be done without much concern for accuracy degradation. FC generation in the visible and near-IR have enjoyed much progress in the last decade however there is still a lack of viable methods for producing FCs in the mid-IR and THz spectrum. Typically, combs are generated with mode-locked pulses or sending coherent light through a non-linear resonator, however there is a lack of materials available to achieve this in the longer wavelengths. Current approaches for FCs at longer wavelength typically include non-linear frequency conversion, which suffers from low conversion efficiency. Quantum Cascade lasers (QCLs) are a ubiquitous source of coherent radiation in the mid-infrared to THz regions of the spectrum. However, passive mode-locking is hard to achieve in QCLs because of their inherently low gain recovery time, due to intersubband operation, on the order of 1ps when compared to the round trip time on the order of 100ps. Despite the difficulty of actively or passively locking QCLs, experimental evidence has shown that frequency combs are indeed generated by free-running QCLs [1,2], i.e. no additional active or passive elements. Given proper dispersion compensation and a broadband gain medium the QCL generates coherent frequency combs with a frequency modulated phase relation.

In an attempt to explicate this behavior, a theoretical, frequency-domain model was developed via a perturbative solution of the Maxwell Bloch equation in the frequency domain[3], illustrating that frequency modulation is indeed a natural consequence of spatial hole burning in an inhomogeneously broadened gain medium (the origin of multi-mode lasing) and a short gain recovery time, which favors constant intensity. This combination of multi-mode operation with constant intensity is a signature of frequency modulation. The theoretical model predicts a pseudo-random frequency modulation of the laser with a general period of oscillations equal to the gain recovery time and an amplitude equal to the gain bandwidth. The work proposed here attempts to explain the necessity of a pseudo-random FM signal as well as take into account spectral hole burning which was excluded in the FD model.

This time domain model is developed using the Optical Bloch Equations (OBE). Initially the question of coherence must be investigated. It is our hypothesis that the dynamics of the laser, in fact, do not rely on coherent processes. In order to prove this, we compare the OBE under a full coherent interaction with a modified rate equation that is an approximation of the OBE. We operate under the assumption that the time rate of change is slower than the loss of coherence. This approximation greatly lightens the computational load, allowing for a more realistic model (more inhomogeneously broadened spectral bins, a longer cavity length, etc.). With the validation of this assumption we assert that the operation regime with the most stimulated emission will result in the lowest threshold, rationalizing the necessity of a pseudo random frequency modulation signal. In order to achieve this, we input various forms of an FM signal into a model utilizing real world specifications of both mid-infrared and THz QCLs. We show that indeed, with an FM signal similar to that produced in the FD model, the gain provided by the active medium peaks for a very random signal that fully spans the gain bandwidth, we trace the root cause of this to be spatial hole burning. Further experimental results have shown that under certain conditions the QCLs exhibit amplitude modulation as well as frequency modulation. In order to keep our model current, we develop analytical solutions for spatial hole burning in the cavity and investigate under what conditions of AM and FM is the spatial hole burning reduced.

Back to top