A Novel Approach to Canonical Divergences within Information Geometry.
Nihat AyShun-ichi AmariPublished in: Entropy (2015)
Keyphrases
- information geometry
- bregman divergences
- fisher information
- information theoretic
- cost sensitive
- maximum entropy
- exponential family
- information theory
- mahalanobis distance
- kl divergence
- learning theory
- nearest neighbor
- loss function
- theoretical guarantees
- special case
- nonnegative matrix factorization
- data dependent
- density estimation
- closed form
- mixture model
- bayesian networks
- negative matrix factorization
- boosting algorithms
- noise level
- multi class
- riemannian metric