Information Geometry of U-Boost and Bregman Divergence.
Noboru MurataTakashi TakenouchiTakafumi KanamoriShinto EguchiPublished in: Neural Comput. (2004)
Keyphrases
- bregman divergences
- cost sensitive
- maximum entropy
- information theoretic
- mahalanobis distance
- nearest neighbor
- loss function
- exponential family
- learning theory
- theoretical guarantees
- special case
- nonnegative matrix factorization
- kl divergence
- data dependent
- multi class
- support vector
- noisy images
- boosting algorithms
- conditional random fields
- mixture model
- text classification
- active learning