Mutual Information Matrices Are Not Always Positive Semidefinite.
Sune K. JakobsenPublished in: IEEE Trans. Inf. Theory (2014)
Keyphrases
- positive semidefinite
- mutual information
- semidefinite programming
- similarity function
- kernel matrix
- metric learning
- linear combination
- similarity measure
- semidefinite
- kernel function
- kernel methods
- positive semidefinite matrices
- image registration
- polynomial kernels
- string kernels
- feature selection
- information gain
- low rank
- support vector
- text categorization
- maximum margin
- linear programming