On the Variance of the Fisher Information for Deep Learning.
Alexander SoenKe SunPublished in: NeurIPS (2021)
Keyphrases
- deep learning
- fisher information
- unsupervised learning
- probability density function
- machine learning
- kl divergence
- positive definite
- estimation error
- euclidean space
- covariance matrix
- geodesic distance
- noise level
- mental models
- standard deviation
- weakly supervised
- em algorithm
- text mining
- free energy
- training data
- similarity measure