Approximate Fisher Information Matrix to Characterise the Training of Deep Neural Networks.

Zhibin Liao Tom Drummond Ian Reid Gustavo Carneiro

Published in: CoRR (2018)

Keyphrases

fisher information
neural network
training process
positive definite
probability density function
euclidean space
information geometry
noise level
multidimensional scaling
geodesic distance
training set
sufficient conditions
error rate
low rank
kullback leibler divergence