Deployment of a Regularized Feature Selection Framework on an Overlay Desktop Grid

Abstract

We present an inexpensive solution to overcome the computational challenges of a state-of-the art feature selection framework. The algorithm relies on two nested loops of cross validation that allow the optimal choice for the regularization parameters resulting in a computationally expensive procedure. In order to speed up the process, we perform the loops on a desktop grid facility deployed on PCs in computer rooms as virtual machines connected through a VPN. The local grid shares resources with similar grids managed by other North Italian universities.