Last Updated: February 25, 2016
·
442
· tedtanne

Know Thy Kernel Methods

When dealing with kernel methods it is paramount to understand the routines as to what you expect to return - e.g. gamma for RBFs otherwise you may find yourself incorrectly using the functions. Also given a set of kernels, there is optimal way (based on feature vector selection), to combine them (sums, products, learned or non learned, with or without cross-validation.