Sparse Linear models¶
A set of generalized linear models of the form \(p(y|Z) = p(y| f(w^Tx))\) for some link function \(f\), where we ecnourage the weight vector w to be sparse. This approach offers significant computational advantages like:
If the number of dimensions D is larger than N. We want to find the smallest subset of features that can accurately predict the response in order to prevent overfitting.
Useful i context of sparse kernel machines.