165
click to vote
NIPS
2004
15 years 4 months ago
2004
The problem of learning a sparse conic combination of kernel functions or kernel matrices for classification or regression can be achieved via the regularization by a block 1-norm...