ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 2111.15256
9
4

Approximate Spectral Decomposition of Fisher Information Matrix for Simple ReLU Networks

30 November 2021
Yoshinari Takeishi
Masazumi Iida
J. Takeuchi
ArXivPDFHTML
Abstract

We argue the Fisher information matrix (FIM) of one hidden layer networks with the ReLU activation function. For a network, let WWW denote the d×pd \times pd×p weight matrix from the ddd-dimensional input to the hidden layer consisting of ppp neurons, and vvv the ppp-dimensional weight vector from the hidden layer to the scalar output. We focus on the FIM of vvv, which we denote as III. Under certain conditions, we characterize the first three clusters of eigenvalues and eigenvectors of the FIM. Specifically, we show that 1) Since III is non-negative owing to the ReLU, the first eigenvalue is the Perron-Frobenius eigenvalue. 2) For the cluster of the next maximum values, the eigenspace is spanned by the row vectors of WWW. 3) The direct sum of the eigenspace of the first eigenvalue and that of the third cluster is spanned by the set of all the vectors obtained as the Hadamard product of any pair of the row vectors of WWW. We confirmed by numerical calculation that the above is approximately correct when the number of hidden nodes is about 10000.

View on arXiv
Comments on this paper