Data-dependence of plateau phenomenon in learning with neural network—statistical mechanical analysis * *This article is an updated version of: Yoshida Y and Okada M 2019 Data-Dependence of plateau phenomenon in learning with neural network–statistical mechanical analysis 33rd Conf. Neural Information Processing Systems (NeurIPS 2019), Vancouver, Canada vol 32 eds H Wallach, H Larochelle, A Beygelzimer, F d‘Alché–Buc, E Fox and

Abstract

The plateau phenomenon, wherein the loss value stops decreasing during the process of learning, has been reported by various researchers. The phenomenon was actively inspected in the 1990s and found to be due to the fundamental hierarchical structure of neural network models. Then, the phenomenon has been thought of as inevitable. However, the phenomenon seldom occurs in the context of recent deep learning. There is a gap between theory and reality. In this paper, using statistical mechanical formulation, we clarified the relationship between the plateau phenomenon and the statistical property of the data learned. It is shown that the data whose covariance has small and dispersed eigenvalues tend to make the plateau phenomenon inconspicuous.