Universal Statistics of Fisher Information in Deep Neural Networks: Mean Field Approach.
Ryo KarakidaShotaro AkahoShun-ichi AmariPublished in: CoRR (2018)
Keyphrases
- fisher information
- neural network
- free energy
- competitive learning
- information geometry
- markov random field
- probability density function
- euclidean space
- belief propagation
- upper bound
- noise level
- belief networks
- closed form
- information theoretic
- self organizing maps
- positive definite
- bayesian inference
- posterior distribution
- exponential family
- model selection
- em algorithm
- graphical models