Hessian Matrix

ConvNets Landscape Convergence: Hessian-Based Analysis of Matricized Networks

This paper introduces a method for estimating the Hessian matrix norm in convolutional neural networks, offering insights into the loss landscape’s local behaviour, supported by empirical convergence analysis.

Unraveling the Hessian: A Key to Smooth Convergence in Loss Function Landscapes

This paper explore the convergence of the loss landscape in neural networks as the sample size increases, focusing on the Hessian matrix to understand the local geometry of the loss function.