TP7a: Learning Compact Representations: Theory and Applications
Tue, 27 Oct, 13:30 - 15:10 PT (UTC -7)
Location: Chapel
Session Type: Lecture
Session Co-Chairs: Shirin Saeedi Bidokhti, University of Pennsylvania, United States and Shirin Jalali, Rutgers University, United States
Track: Adaptive Systems, Machine Learning, and Data Analytics
Tue, 27 Oct, 13:30 - 13:55 PT (UTC -7)

TP7a.1: Unsupervised Audio Denoising via Neural Lossy Compression
Aayush Rajesh, Tsachy Weissman, Stanford University, United States
Tue, 27 Oct, 13:55 - 14:20 PT (UTC -7)

TP7a.2: Communications for Vision-Language Models: Task-Aware Visual Token Transmission
Harshithanjani Athi, Sravan Kumar Ankireddy, The University of Texas at Austin, United States; Jianzhong Charlie Zhang, Samsung Research America, Samsung, United States; Hyeji Kim, The University of Texas at Austin, United States
Tue, 27 Oct, 14:20 - 14:45 PT (UTC -7)

TP7a.3: Unconditional CNN denoisers contain sparse semantic representation of images
Zahra Kadkhodaie, Massachusetts Institute of Technology, United States; Stéphane Mallat, Collège de France, France; Eero Simoncelli, New York University, United States
Tue, 27 Oct, 14:45 - 15:10 PT (UTC -7)

TP7a.4: On the Optimality of Neural Compression for Structured Stationary Sources
Ali Zafari, Rutgers University, United States; Jinheng Zhang, Shirin Saeedi Bidokhti, University of Pennsylvania, United States; Shirin Jalali, Rutgers University, United States