Fast Simultaneous Training of Generalized Linear Models (FaSTGLZ)

TL;DR

The paper introduces FaSTGLZ, an efficient algorithm for simultaneously training multiple sparse generalized linear models, significantly reducing computation time for tasks like bootstrapping and permutation testing.

Contribution

It presents a novel algorithm that exploits redundancies across related problems to enable fast simultaneous training of generalized linear models.

Findings

Achieves significant computational speedups over sequential methods.

Successfully applied to real-world datasets for bootstrapping and permutation testing.

Facilitates statistically rigorous classification with reduced computational resources.

Abstract

We present an efficient algorithm for simultaneously training sparse generalized linear models across many related problems, which may arise from bootstrapping, cross-validation and nonparametric permutation testing. Our approach leverages the redundancies across problems to obtain significant computational improvements relative to solving the problems sequentially by a conventional algorithm. We demonstrate our fast simultaneous training of generalized linear models (FaSTGLZ) algorithm on a number of real-world datasets, and we run otherwise computationally intensive bootstrapping and permutation test analyses that are typically necessary for obtaining statistically rigorous classification results and meaningful interpretation. Code is freely available at http://liinc.bme.columbia.edu/fastglz.