Learning Taxonomies by Dependence Maximization
Matthew B. Blaschko & Arthur Gretton Max Planck Institute for Biological Cybernetics -

Poster ID M8

Simultaneously cluster data and find a taxonomy that relates the clusters Makes use of the well studied field of numerical taxonomy Maximize a kernel dependence measure, HSIC, between the data and taxonomy Efficient optimization with spectral relaxation

Faces clustered by identity and expression
reinforcement learning Bayesian learning

discriminative learning

neuroscience

neural network applications

miscelaneous

neural network training

hardware

NIPS papers clustered by topic