NIPS
2004
13 years 5 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...