Technical Program

Paper Detail

Paper Title Bio-inspired problems in rate-distortion theory
Paper IdentifierWE2.R1.2
Authors Sarah Marzen, Massachusetts Institute of Technology, United States
Session Deep Learning for Compression
Location Le Théatre (Parterre), Level -1
Session Time Wednesday, 10 July, 11:40 - 13:20
Presentation Time Wednesday, 10 July, 12:00 - 12:20
Abstract Some biologists believe that sensory systems have evolved to near-optimally and lossily compress sensory stimuli. Rate-distortion theory can, therefore, be used to benchmark the performance of sensory systems. I will first explore constraints on sensory systems that can adapt to optimally lossily compress signals from different environments. I will then explore how well a retina-inspired image autoencoder can lossily compress natural images.