CSE 656 Seminars in Computer Vision - Wednesdays 11:30am-12:50pm, Room NCS 120

The overall purpose of this seminar is to bring together people with interests in Computer Vision theory and techniques and to examine current research issues. This course will be appropriate for people who already took a Computer Vision graduate course or already had research experience in Computer Vision. To enroll in this course, you must either: (1) be in the PhD program or (2) receive permission from the instructors.

Each seminar will consist of multiple short talks (around 10 minutes) by multiple people. Students can register for 1 credit for CSE656. Registered students must attend and present a minimum of 2 or 3 talks. Everyone else is welcome to attend. Fill in https://forms.gle/pCVXovgfMfQwGqG38 to subscribe to our mailing list for further announcement.
Title: Class visual similarity based noisy sample removal in generative Few Shot Learning
Time: Thursday, Feb 4, 11:30am - 1:00pm
Zoom:
https://stonybrook.zoom.us/j/8563646526?pwd=anJna1gzUStXNlNVSUIzdDRUSC9CUT09

Meeting ID: 856 364 6526
Passcode: 203791



Abstract:  

Over the past decade, larger datasets, hardware accelerations, and network architecture improvements have contributed to phenomenal achievements in many tasks of computer

vision. However, in the absence of large datasets, computer vision models struggle to learn

general representations which results in poor performance. Few-shot learning tries to address 

this problem by proposing models which learn from a few examples.


I first give an overall review of few-shot learning methods. I particularly focus on generative Few Shot Learning(FSL) methods, which augment the scarce categories in a dataset by generating samples for those rare categories. As the actual class distribution can be complex and lie very close to each other, the sample generated for one class can be noisy or lie close to another class.  However, none of the current FS generative methods perform any form of quality control of the generated samples.


In this work, I propose to identify and remove the generated samples that are less likely to be in the distribution of the few-shot class. Here I particularly deal with few-shot scenarios where the

prior information of the relationship between the classes based on visual  similarity is available. The main idea is to exploit these priors to better identify the unreliable generated samples.


Particularly, I have proposed two methods based on class relationship to detect noisy generated samples. In the first method, we assume that the embedding space of each class follows a Gaussian distribution.  From this assumption, I propose Gaussian Neighborhood (GN), a method to estimate how likely a generated sample is drawn from the estimated distribution of a few-shot class.  We evaluate this method on the Hematopoiesis dataset.  By simply eliminating samples based on thresholding our proposed GN scores, the few-shot  classification  performance  is  improved by 5% and 2% in five shot and one shot respectively, compared to the model trained on all generated images. 


The GN scores represent the similarity distances from the generated samples to their classes, based on the assumption that each class is a Gaussian distribution.  However, this assumption might be strict in many scenarios since the real distributions of data can be arbitrarily complex.  Thus in my second proposed method, I aim to learn such similarity distances directly from data via metric learning. I propose to train a deep-network to regress the similarity distance between a pair of samples.  This network is trained using both the class-level  visual  similarity  information  and  the  class  labels.   This method improves the 1-shot and 5-shot classification performances by 0.5% and 1% respectively, compared to GN.

Abstract: Generative image models trained on massive datasets encode the statistics of our visual world. An off-the-shelf diffusion or flow-matching model can therefore serve as a general-purpose image prior, providing information about images across a wide range of tasks. Harnessing these capabilities, however, requires combining the model at inference time with additional signals or constraints unknown during training. This thesis will develop training-free methods for inference under a generative denoising prior. I first show how sampling from a trained denoiser can be formulated as an optimization problem, which combined with a constraint, can solve tasks ranging from conditional generation and weakly supervised segmentation to combinatorial optimization. I then demonstrate how the inherent properties of denoisers can accelerate inference under such constraints. Replacing gradient descent with an inexact Newton update, based on the symmetry of the denoiser's Jacobian, substantially reduces inference costs without tradeoffs. I also explore a middle ground between model adaptation and fully training-free inference by using the denoiser's robust internal representations to learn constraints from limited labeled data. These methods are applied to gigapixel image domains such as digital histopathology and remote sensing, where generative models can only be trained at a patch scale. To synthesize arbitrarily large images at resolutions unseen during training, I introduce inference-time algorithms that enforce consistency across spatially overlapping patches and image scales. Finally, I propose repurposing inference-time algorithms from sampling tools, to mechanisms for understanding the prior learned by a denoising model. Extending the previously developed techniques, I analyze the Jacobians of generative denoisers, where preliminary results indicate that Jacobian spectra correlates with generative quality. This motivates a Jacobian-spectrum regularization as a way to improve model performance using insights derived from inference-time algorithms.

Speaker: Alexandros Graikos

Location: NCS 220
Making sense of Twitter @ Bloomberg presented by Daniel Preotiuc-Pietro

ABSTRACT: The Bloomberg Terminal has provided ways for investors and journalists to sift through and understand the immense volume of tweets and discover financially-relevant content ever since the SEC approved the use of Twitter for company disclosures back in 2013.

In the first part of the talk, I will showcase how tweets impact financial markets and how Bloomberg is using Natural Language Processing methods to identify financially relevant tweets that move the markets. Our processing pipeline feeds directly to clients, journalists in the newsroom and powers several news analytic products offered by the company including trending companies and consumer sentiment for publicly traded equities.

However, understanding user pragmatic intent in individual tweets would allow us to gain deeper insights and enable new applications. I will present several recent research studies focused on understanding intent including identifying complaints and the roles with which vulgarity is used in social media and how these can help improve applications such as sentiment analysis and hate speech detection.

BIO: Daniel Preotiuc-Pietro is a Senior Research Engineer and Team Lead at Bloomberg LP, where he works on analyzing and building models for real-world large scale social media and news mining and information extraction. His research interests are focused on understanding the social and temporal aspects of text, especially from social media, with applications in domains such as Social Psychology, Law, Political Science and Journalism. Several of his research studies were featured in popular press including the Washington Post, BBC, New Scientist, Scientific American or FiveThirtyEight. He is a co-organizer of the Natural Legal Language Processing workshop series. Prior to joining Bloomberg LP, Daniel was a postdoctoral researcher at the University of Pennsylvania with the interdisciplinary World Well Being Project and obtained his PhD in Natural Language Processing and Machine Learning at the University of Sheffield, UK.
Stony Brook Radiology Department Invites you to the AI Research Seminar


Abstract: This talk traces the evolution of AI in neuroradiology through three connected stages: Precision Care, General Medical Intelligence, and Discovery. It moves from multimodal deep learning for stroke outcome prediction, to large-scale foundation models and agent-based medical AI, to imaging biomarkers such as brain aging derived from routine scans. It closes with early-stage work on Galen, a framework for reproducible, scalable, and transparent biomedical imaging AI development and analysis.

Speaker: Yongkai Liu, PhD, Instructor, Department of Radiology (Neuroimaging & Neurointervention), Stanford University

Join Microsoft Teams: https://teams.microsoft.com/meet/221225483255320?p=Anban9lwgOgAq3jaHj
Meeting ID: 221 225 483 255 320
Passcode: kJ7fF3ou
Abstract: Artificial intelligence (AI) is rapidly transforming scientific discovery, enabling breakthroughs in areas ranging from drug discovery to modeling complex physical systems. In the life sciences, AI has traditionally been applied to prediction tasks such as classifying molecules as toxic or non-toxic, estimating drug properties, or solving partial differential equations. These discriminative models have proven powerful, but they are inherently limited to mapping existing inputs to deterministic outputs. A new wave of methods is shifting the paradigm from discrimination to generation: creating new possibilities, such as generating novel molecules or designing new drugs. By reframing AI as both a predictive and generative engine, this shift offers new pathways for accelerating discovery and innovation in life sciences at an unprecedented scale. This talk will cover several aspects of AI for Science (AI4Sci), beginning with advances in discriminative models for molecular systems and solving PDEs, and then turning to generative approaches, including diffusion models for 3D molecular generation and large language models for drug editing. Together, these developments illustrate how moving from prediction to creation is redefining what AI can contribute to science.

Bio: Wenhan Gao is a fourth-year Ph.D. student in Applied Mathematics under the supervision of Professor Yi Liu. He was also a Staff Research Scientist Intern at VISA Research, where he worked on large language models (LLMs) and multi-agent systems for commerce. Wenhan's research focuses on AI for Science (AI4Sci), with a particular emphasis on generative AI. His work looks deep into the fundamental mechanisms of AI models when applied to scientific tasks, and he strives to incorporate established scientific priors, such as symmetry, into model design. He has published papers as a first or corresponding author in leading AI and computational venues, including ICLR, ICML, NeurIPS, TMLR, ACL, and the Journal of Computational Physics. In addition to his research, Wenhan has served as a reviewer and oral session chair for top AI conferences and as a lecturer for both undergraduate and graduate courses at Stony Brook University.

Location: IACS Seminar Room or Zoom

This seminar will take place in person and online*

Join Zoom Meeting: https://stonybrook.zoom.us/j/91670093552?pwd=2EcniXqPZLTpa4ZBKRs1zAjYqs1LS0.1

Meeting ID: 916 7009 3552
Passcode: 434045

Yihui Ren

AI Department, Brookhaven National Laboratory

Abstract: Artificial intelligence is rapidly changing how we approach scientific discovery and engineering design. However, realizing the potential of modern AI in scientific domains requires going beyond simply applying models developed for natural language: scientific data have distinct structures, and scientific workflows often require tight integration among learned representations, domain knowledge, simulation, and numerical optimization.
In this talk, I will present two recent efforts that explore complementary directions for building AI systems for science and engineering. First, FM4NPP investigates whether the foundation-model paradigm can scale to experimental nuclear and particle physics. Using more than 11 million particle-collision events, we develop a self-supervised learning framework for sparse detector data and demonstrate scaling up to 188 million parameters. The resulting representations generalize across diverse downstream tasks and enable data-efficient adaptation with lightweight task-specific components.
Second, I will present AutoSizer, an LLM-driven framework for automatic sizing of analog and mixed-signal circuits. Rather than asking an LLM to perform numerical optimization directly, AutoSizer uses the LLM as a reasoning and orchestration layer that interprets circuit behavior, analyzes simulation feedback, and adaptively refines the optimization process.
Together, these studies illustrate two complementary roles for modern AI in scientific workflows: learning reusable representations from complex scientific data and reasoning over computational tools to guide challenging design and optimization problems. I will discuss the lessons from these efforts and opportunities toward more general, scalable, and reusable AI systems for science and engineering.

Short bio: Dr. Yihui (Ray) Ren is a Senior Computational Scientist in AI/ML and the AI-Codesign Group Lead in the Artificial Intelligence Department within Brookhaven National Laboratory's Computing and Data Sciences Directorate. His research focuses on AI for Science, AI-hardware codesign, real-time AI systems, and the evaluation of emerging AI hardware architectures. He also develops advanced AI methodologies for nuclear safeguards applications. Dr. Ren's key contributions include unpaired domain mapping techniques to mitigate domain shift, generative surrogate modeling, object detection and video synthesis using event-based cameras, and the development of AI foundation models such as the Foundation Model for Nuclear and Particle Physics (FM4NPP). Prior to joining Brookhaven National Laboratory in 2018, he was a postdoctoral researcher at Virginia Tech. He received his Ph.D. in Physics from the University of Notre Dame in 2015.

Location: NCS 120

Zoom: https://stonybrook.zoom.us/j/92507585369?pwd=9bw2hsQmLbladVpPRiaWK94gYlwCGQ.1&jst=2

The Pittsburgh Supercomputing Center is pleased to present a Machine Learning and Big Data workshop.

This workshop will focus on topics including big data analytics and machine learning with Spark, as well as deep learning.

This will be an IN PERSON event hosted by various satellite sites, there WILL NOT be a direct to desktop option for this event. SBU's Institute for Advanced Computational Science (IACS) is one of those satellite sites!

Location: IACS Conference Room #2

Interested applicants must first have an ACCESS ID. If you don't have the ID, please visit this page to create one: ACCESS USER REGISTRATION.


Once you have an ACCESS ID, please login (see top right here) then register here.