Yes, my deep network works! But.. what did it learn?
| Quand ? |
Le 14/05/2024, de 15:00 à 16:00 |
|---|---|
| Participants |
Jeremias Sulam |
| Ajouter un événement au calendrier |
vCal
iCal
|
Jeremias Sulam
Title: Yes, my deep network works! But.. what did it learn?
Abstract: Modern machine learning methods are revolutionizing what we can do with data — from TikTok video recommendations to biomarker discovery in cancer research. Yet, the complexity of these deep models makes it harder to understand what functions these data-dependent models are computing, and which features they learn to be important for a given task. In this talk, I will review two approaches for turning general deep learning models more interpretable. The first one will study an unsupervised setting in the context of imaging inverse problems and will show how to design and train networks that provide exact proximal operators that approximate that of the (log) prior distribution of the data. The second will switch to supervised classification problems for computer vision, where will re-think the use of Shapley coefficients for black-box model explanations.
Bio: Jeremias Sulam received his bioengineering degree from Universidad Nacional de Entre Ríos, Argentina, in 2013, and his PhD in Computer Science from the Technion – Israel Institute of Technology, in 2018. He joined the Biomedical Engineering Department at Johns Hopkins University in 2018 as an assistant professor, and he is also a core faculty at the Mathematical Institute for Data Science (MINDS) and the Center for Imaging Science at JHU. He is the recipient of the Best Graduates Award of the Argentinean National Academy of Engineering, and the Early CAREER award of the National Science Foundation. His research interests include inverse problems, sparse representation modeling and machine learning.
Website: https://sites.google.com/view/jsulam
In Room M7.101 of Monod campus, ENSL.