Phase Diagram of Stochastic Gradient Descent in High-Dimensional Two-Layer Neural Networks

Explore the phase diagram of Stochastic Gradient Descent in high-dimensional two-layer neural networks in this 18-minute workshop talk presented by B. Loureiro from EPFL. Delve into the intricacies of neural network optimization as part of the Workshop on Spin Glasses, organized by NCCR SwissMAP. Gain insights into the behavior of SGD in complex neural architectures and its implications for machine learning research and applications.

Phase diagram of Stochastic Gradient Descent in high-dimensional two-layer neural networks