{"id":54151,"date":"2025-09-29T11:52:32","date_gmt":"2025-09-29T15:52:32","guid":{"rendered":"https:\/\/engineering.jhu.edu\/ams\/?post_type=news&p=54151"},"modified":"2025-10-01T09:32:38","modified_gmt":"2025-10-01T13:32:38","slug":"the-secret-geometry-hidden-inside-every-image","status":"publish","type":"news","link":"https:\/\/engineering.jhu.edu\/ams\/news\/the-secret-geometry-hidden-inside-every-image\/","title":{"rendered":"The secret geometry hidden inside every image"},"content":{"rendered":"

Johns Hopkins researchers have developed a new framework that uses graph neural networks (GNNs) to classify images through geometry\u2014an advance that could help improve how machines flag inappropriate content and identify diseases in medical images. GNNs are a type of AI that learns by mapping relationships between data points. <\/span>Their method promises better performance when working with limited data by leveraging the hidden structure, or manifold, underneath data points like images.\u00a0<\/span>\u00a0<\/span><\/p>\n

The team\u2019s approach, which appears on the preprint site <\/span>ArXiv<\/span><\/a>, <\/span>captures the hidden geometric relationships in data that help models make more accurate and generalizable decisions. Even better, the researchers say the new method is simple enough to integrate into existing systems without requiring a major overhaul.<\/span>\u00a0<\/span><\/p>\n

\u201cWe know that graph neural networks are powerful tools, especially when working with data that\u2019s naturally connected\u2014like social networks or route maps,\u201d said Caio F. Deberaldini Netto, a PhD student in <\/span>applied mathematics and statistics<\/span><\/a> at the Whiting School of Engineering<\/span> \ufffc<\/span>and lead author of the study. \u201c<\/span>But arbitrary data, like images\u2013which can be seen as pixel grids or language, which can be seen as token sequences\u2013don’t naturally look like graphs<\/span>.<\/span>\u00a0<\/span><\/p>\n

The trick, the researchers say, is to find the manifold\u2014a hidden, low-dimensional geometric pattern that reveals how seemingly different images are actually related. This idea, known as the manifold hypothesis, suggests that although image data is high-dimensional (like millions of pixels), the real patterns that matter\u2014like \u201cthis is a cat\u201d\u2014live in a much simpler structure.<\/span>\u00a0<\/span><\/p>\n

Using a machine learning model called a variational autoencoder (VAE), the team first mapped images into this lower-dimensional manifold. Then, they constructed a graph where each image was a node, and connections were made based on how close the images were in this geometric space.<\/span>\u00a0<\/span><\/p>\n

\u201cIf two pictures of dogs are close together in the manifold\u2014even if the model doesn\u2019t know they\u2019re dogs\u2014it still picks up the similarity based on features like shape or color,\u201d explained co-author <\/span>Luana Ruiz<\/span><\/a>, assistant professor of applied mathematics and statistics.<\/span>\u00a0<\/span><\/p>\n

Once the graph of images is built, a GNN processes it by learning from these connections\u2014propagating information between similar images. This allows the model to learn patterns not just from individual images, but also from the relationships between them.<\/span>\u00a0<\/span><\/p>\n

The team found that across multiple datasets\u2014including CIFAR-10 (common objects), CelebA (celebrity faces), and even medical images\u2014the method consistently outperformed other strategies like the most basic type of neural network (called \u201cvanilla\u201d)<\/span> or<\/span> superpixel-based graphs, which are networks of visually similar pixel groups. Most importantly, it showed better generalization: the ability to correctly classify new, unseen images.<\/span>\u00a0<\/span><\/p>\n

\u201cWe tested not just on new images, but also on larger and more complex graphs,\u201d said Deberaldini Netto. \u201cOur GNN\u2019s generalization gap\u2014the difference between training and test accuracy\u2014shrinks as you add more data. That\u2019s exactly what you want.\u201d<\/span>\u00a0<\/span><\/p>\n

The researchers say this approach is ready to be plugged into real-world pipelines\u2014from medical diagnosis to content recommendation\u2014anywhere models need to make robust decisions based on complex data.<\/span>\u00a0<\/span><\/p>\n

\u201cSay you’re classifying patient scans,\u201d Deberaldini Netto explained. \u201cYou can\u2019t just rely on pixel values. You need to learn from patterns across similar patients. Our method lets you build those patterns into the model from the start.\u201d<\/span>\u00a0<\/span><\/p>\n

Additionally, the researchers are already applying it to textual data\u2014like authorship attribution, where relational structure between written passages could help determine whether a work belongs to Shakespeare or Marlowe. And because it\u2019s compatible with existing models, like large language models or convolutional networks, it could become a powerful building block in general-purpose AI, they say.<\/span>\u00a0<\/span><\/p>\n

\u201cThis work is about making models that are not just accurate, but also smarter and more adaptable,\u201d said Ruiz. \u201cIt\u2019s a step toward AI that really understands structure in the world\u2014whether that\u2019s in pixels, words, or people.\u201d<\/span>\u00a0<\/span><\/p>\n","protected":false},"template":"","class_list":["post-54151","news","type-news","status-publish","hentry","news_categories-applied-mathematics","news_categories-research"],"acf":[],"yoast_head":"\nThe secret geometry hidden inside every image | Department of Applied Mathematics and Statistics<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/engineering.jhu.edu\/ams\/news\/the-secret-geometry-hidden-inside-every-image\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"The secret geometry hidden inside every image | Department of Applied Mathematics and Statistics\" \/>\n<meta property=\"og:description\" content=\"Johns Hopkins researchers have developed a new framework that uses graph neural networks (GNNs) to classify images through geometry\u2014an advance that could help improve how machines flag inappropriate content and…\" \/>\n<meta property=\"og:url\" content=\"https:\/\/engineering.jhu.edu\/ams\/news\/the-secret-geometry-hidden-inside-every-image\/\" \/>\n<meta property=\"og:site_name\" content=\"Department of Applied Mathematics and Statistics\" \/>\n<meta property=\"article:modified_time\" content=\"2025-10-01T13:32:38+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/engineering.jhu.edu\/ams\/wp-content\/uploads\/2025\/09\/AdobeStock_605382790-1024x492.jpeg\" \/>\n\t<meta property=\"og:image:width\" content=\"1024\" \/>\n\t<meta property=\"og:image:height\" content=\"492\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:image\" content=\"https:\/\/engineering.jhu.edu\/ams\/wp-content\/uploads\/2025\/09\/AdobeStock_605382790.jpeg\" \/>\n<meta name=\"twitter:label1\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data1\" content=\"3 minutes\" \/>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"The secret geometry hidden inside every image | Department of Applied Mathematics and Statistics","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/engineering.jhu.edu\/ams\/news\/the-secret-geometry-hidden-inside-every-image\/","og_locale":"en_US","og_type":"article","og_title":"The secret geometry hidden inside every image | Department of Applied Mathematics and Statistics","og_description":"Johns Hopkins researchers have developed a new framework that uses graph neural networks (GNNs) to classify images through geometry\u2014an advance that could help improve how machines flag inappropriate content and…","og_url":"https:\/\/engineering.jhu.edu\/ams\/news\/the-secret-geometry-hidden-inside-every-image\/","og_site_name":"Department of Applied Mathematics and Statistics","article_modified_time":"2025-10-01T13:32:38+00:00","og_image":[{"width":1024,"height":492,"url":"https:\/\/engineering.jhu.edu\/ams\/wp-content\/uploads\/2025\/09\/AdobeStock_605382790-1024x492.jpeg","type":"image\/jpeg"}],"twitter_card":"summary_large_image","twitter_image":"https:\/\/engineering.jhu.edu\/ams\/wp-content\/uploads\/2025\/09\/AdobeStock_605382790.jpeg","twitter_misc":{"Est. reading time":"3 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/engineering.jhu.edu\/ams\/news\/the-secret-geometry-hidden-inside-every-image\/","url":"https:\/\/engineering.jhu.edu\/ams\/news\/the-secret-geometry-hidden-inside-every-image\/","name":"The secret geometry hidden inside every image | Department of Applied Mathematics and Statistics","isPartOf":{"@id":"https:\/\/engineering.jhu.edu\/ams\/#website"},"datePublished":"2025-09-29T15:52:32+00:00","dateModified":"2025-10-01T13:32:38+00:00","breadcrumb":{"@id":"https:\/\/engineering.jhu.edu\/ams\/news\/the-secret-geometry-hidden-inside-every-image\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/engineering.jhu.edu\/ams\/news\/the-secret-geometry-hidden-inside-every-image\/"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/engineering.jhu.edu\/ams\/news\/the-secret-geometry-hidden-inside-every-image\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/engineering.jhu.edu\/ams\/"},{"@type":"ListItem","position":2,"name":"News","item":"https:\/\/engineering.jhu.edu\/ams\/news\/"},{"@type":"ListItem","position":3,"name":"The secret geometry hidden inside every image"}]},{"@type":"WebSite","@id":"https:\/\/engineering.jhu.edu\/ams\/#website","url":"https:\/\/engineering.jhu.edu\/ams\/","name":"Hopkins Applied Math & Statistics","description":"Department of Applied Mathematics and Statistics","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/engineering.jhu.edu\/ams\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"}]}},"distributor_meta":false,"distributor_terms":false,"distributor_media":false,"distributor_original_site_name":"Department of Applied Mathematics and Statistics","distributor_original_site_url":"https:\/\/engineering.jhu.edu\/ams","push-errors":false,"_links":{"self":[{"href":"https:\/\/engineering.jhu.edu\/ams\/wp-json\/wp\/v2\/news\/54151","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/engineering.jhu.edu\/ams\/wp-json\/wp\/v2\/news"}],"about":[{"href":"https:\/\/engineering.jhu.edu\/ams\/wp-json\/wp\/v2\/types\/news"}],"wp:attachment":[{"href":"https:\/\/engineering.jhu.edu\/ams\/wp-json\/wp\/v2\/media?parent=54151"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}