Why Deep Learning Works: Physics, Symmetry and Compression
MITCBMM · Talk · 52 min · 2016 · YouTube title: Connections between physics and deep learning
Why it's here
HN recommends it as a clear argument for why deep and cheap learning succeeds, often alongside the associated paper.
What it covers
Max Tegmark's MIT CBMM talk on why deep learning works so well. He argues that the laws of physics and the symmetries in physical data make it compressible into small neural networks, tying physics to network structure.