Fisher Information Matrix

Creator
Creator
Seonglae Cho
Created
Created
2023 Nov 25 11:44
Editor
Edited
Edited
2025 Mar 19 22:8
Refs
Refs
Informant

FIM

Def:
Variance
of
Informant
. FIM reflects
Curvature
of probability distribution since informant indicates the sensitivity to changes in parameter values.

Fisher (Information) Matrix

Def: Expectation for
Outer Product
of score function. If score function (
Informant
) is vector, Fisher matrix acts like
Covariance Matrix
(not identical but equal when average of score function is 0 mathematically).
F=E[s(θ)s(θ)T]F = \mathbb{E}[s(\theta)s(\theta)^T]

Empirical Fisher Matrix (EFIM)

F^=1Ns(θi)s(θi)T\hat{F} = \frac{1}{N}s(\theta_i)s(\theta_i)^T

Hessian Matrix

When the loss function is negative log-likelihood (NLL) or cross entropy, and optimization has converged using likelihood, the Hessian Matrix of loss function HFH \approx F approximates the Fisher Information Matrix.
 
 
 
 
 

Recommendations