Eigenvalue-corrected Natural Gradient Based on a New Approximation

TL;DR

This paper introduces TEKFAC, a novel second-order optimization method for deep neural networks that combines eigenvalue correction, a new Fisher information matrix approximation, and damping techniques, leading to improved training performance.

Contribution

The paper proposes TEKFAC, integrating eigenvalue correction with a new Fisher approximation and damping, advancing second-order optimization methods for DNN training.

Findings

TEKFAC outperforms SGD with momentum, Adam, EKFAC, and TKFAC in experiments.

The method effectively corrects re-scaling factors in the eigenbasis.

Empirical results show improved convergence and training efficiency.

Abstract

Using second-order optimization methods for training deep neural networks (DNNs) has attracted many researchers. A recently proposed method, Eigenvalue-corrected Kronecker Factorization (EKFAC) (George et al., 2018), proposes an interpretation of viewing natural gradient update as a diagonal method, and corrects the inaccurate re-scaling factor in the Kronecker-factored eigenbasis. Gao et al. (2020) considers a new approximation to the natural gradient, which approximates the Fisher information matrix (FIM) to a constant multiplied by the Kronecker product of two matrices and keeps the trace equal before and after the approximation. In this work, we combine the ideas of these two methods and propose Trace-restricted Eigenvalue-corrected Kronecker Factorization (TEKFAC). The proposed method not only corrects the inexact re-scaling factor under the Kronecker-factored eigenbasis, but also considers the new approximation method and the effective damping technique proposed in Gao et al. (2020). We also discuss the differences and relationships among the Kronecker-factored approximations. Empirically, our method outperforms SGD with momentum, Adam, EKFAC and TKFAC on several DNNs.