Technical Program

Paper Detail

Paper Title Panel Discussion: Deep Learning for Compression
Paper IdentifierWE2.R1.5
Authors Aaron B. Wagner, Cornell University, United States; Lav Varshney, University of Illinois at Urbana-Champaign, 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, 13:00 - 13:20
Abstract There is a close link between data compression and generative modeling: decompressors can act as generative models, and inverting a generative model yields a compressor. Recently, deep neural networks, especially in the form of GANs (generative adversarial networks) and VAEs (variational autoencoder), have enabled significant advances in producing generative models for complicated sources such as images. Yet the implications of these advances for compression are only beginning to be explored. This session will focus on recent advances in using deep neural networks to produce generative models with a particular focus on applications to multimedia data compression.