SKETCH TOWARD ONLINE RISK MINIMIZATION

Date
2022-01-27
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract

Empirical risk minimization (ERM) is perhaps the most in uential idea in statistical learning, with applications to nearly all scienti c and technical domains in the form of regression and classi cation models. The growing concerns about the high energy cost of training and the increased prevalence of massive streaming datasets have led many ML practitioners to look for approximate ERM models that can achieve low cost on memory and latency for training. To this end, we propose STORM- Sketch Toward Online Risk Minimization, an online sketching-based method for empirical risk minimization. STORM compresses a data stream into a tiny array of integer counters. This sketch is su cient to estimate a variety of surrogate losses over the original dataset. We provide rigorous theoretical analysis and show that STORM can estimate a carefully chosen surrogate loss for regularized least-squares regression and a margin loss for classi cation. We perform an exhaustive experimental comparison for regression and classi cation training on real-world datasets, achieving an approximate solution with size even less than a data sample.

Description
Degree
Master of Science
Type
Thesis
Keywords
Empirical risk minimization, Sketching
Citation

Gupta, Gaurav. "SKETCH TOWARD ONLINE RISK MINIMIZATION." (2022) Master’s Thesis, Rice University. https://hdl.handle.net/1911/113483.

Has part(s)
Forms part of
Published Version
Rights
Copyright is held by the author, unless otherwise indicated. Permission to reuse, publish, or reproduce the work beyond the bounds of fair use or other exemptions to copyright law must be obtained from the copyright holder.
Link to license
Citable link to this page