Sanghani-students

Download

sanghani-students@cs.vt.edu

October 2024

2 participants
2 discussions

Talk on ML interpretability
by Jia, Ruoxi 21 Oct '24

21 Oct '24

Hi all, As part of our ECE 6514 trustworthy ML course, we have invited an external speaker to give a guest lecture on ML interpretability. If you or any of your students or collaborators are interested in this area, we’d be happy to have you stop by! Bio: Dr. Ian Covert is a postdoctoral researcher at Stanford University, where I'm advised by James Zou and Tatsu Hashimoto. He was previously a Ph.D. student at the University of Washington, where I was advised by Su-In Lee. He also worked with Emily Fox at UW, He was a student researcher at Google Brain working with Jiening Zhan and Ming Jack Po, and I interned as a quantitative researcher at Citadel Securities and Goldman Sachs. Before that, He completed my Bachelor's degree at Columbia University, where he did computational neuroscience research in Liam Paninski's lab. He co-created an explainable AI course in Spring 2022. Title: Interpreting Black-Box Models: Shapley Values and Beyond Time: 10/22/2024 Tuesday 11am-12:15 Abstract: Modern machine learning relies heavily on black-box models, including transformers, CNNs, and tree ensembles, depending on the application. These models provide state-of-the-art accuracy but make it difficult to understand the logic that drives their predictions. This talk discusses “feature attribution” approaches that quantify the influence of each input feature, and provide one of the main lenses to interpret model predictions. We first cover removal-based explanations, a broad class of algorithms encompassing many of the most popular approaches in the field (LIME, SHAP, permutation tests, etc.); we then focus on Shapley values, including their origins in game theory and techniques to accelerate their computation. Zoom: Meeting ID: 840 8576 5484 Passcode: 887334 Best, Ruoxi -- Ruoxi Jia, Ph.D. ruoxijia.net<https://ruoxijia.net/> Assistant Professor Bradley Department of Electrical and Computer Engineering Virginia Tech

1 0

Fw: Nov 12 2pm Distinguished lecture by Aidong Zhang (UVa)
by Shalaby, Wanawsha 11 Oct '24

11 Oct '24

Please see attached flyer. Wanawsha ________________________________ From: Yao, Danfeng via Nvcstaff <nvcstaff(a)cs.vt.edu> Sent: Friday, October 11, 2024 12:02 PM To: CS Faculty & Staff <facstaff(a)cs.vt.edu> Subject: [Nvcstaff] Nov 12 2pm Distinguished lecture by Aidong Zhang (UVa) Good afternoon! I am happy to announce a Distinguished Lecture jointly hosted by the Sanghani Center for Artificial Intelligence and Data Analytics and the CS department. Speaker: Dr. Aidong Zhang, Thomas M. Linville Professor of Computer Science at UVa Title: Can large language models help with scientific hypothesis generation? Time: Nov 12th 2pm-3:15pm Location: * Kelly 310 in Blacksburg * or watch over Zoom in Room 3-024 in VTRC-A, Arlington Please circulate to your students. See the attached flyer for more information. Best, Daphne

1 0