Approximate Coding for Matrix Multiplication in Optimization and Learning

Explore a 32-minute lecture on approximate coding computing presented by Stark Draper from the University of Toronto at the Simons Institute. Delve into the concept of coded matrix multiplication for scenarios where approximate results are sufficient, with potential applications in optimization and learning. Examine the development of rate-distortion analogs for coded computing, inspired by recent research on "epsilon-coded-computing." Discover how to implement schemes with multiple stages of recovery, introducing the concept of "successive approximation" coding as an analog to successive refinement in rate distortion theory.