Hessian Eigenvectors and Principal Component Analysis of Neural Network Weight Matrices

David Haink

This study delves into the intricate dynamics of trained deep neural networks and their relationships with network parameters. Trained networks predominantly continue training in a…