conftrace_

2013 ICML ICML 2013

Learning Optimally Sparse Support Vector Machines

Abstract

We show how to train SVMs with an optimal guarantee on the number of support vectors (up to constants), and with sample complexity and training runtime bounds matching the best known for kernel SVM optimization (i.e. without any additional asymptotic cost beyond standard SVM training). Our method is simple to implement and works well in practice.

🚀 Conference Pioneer - ICML 2013

🌉 Interdisciplinary Bridge - Artificial Intelligence and Mathematics & Optimization

Authors

Andrew Cotter, Shai Shalev-shwartz, Nati Srebro

Topics

Machine Learning > Core Methods > Classification Machine Learning > Optimization & Theory > Optimization Mathematics & Optimization > Optimization > Sparse Optimization Machine Learning > Core Methods > Support Vector Machine

Keywords

feature selection sparse optimization support vector machine generalization guarantee sparse classification kernel methods

More from ICML 2013

All ICML 2013 papers →

Principal Component Analysis on non-Gaussian Dependent Data

Learning Linear Bayesian Networks with Latent Variables

Multiple Identifications in Multi-Armed Bandits

Dynamic Probabilistic Models for Latent Feature Propagation in Social Networks

Efficient Sparse Group Feature Selection via Nonconvex Optimization

Domain Adaptation for Sequence Labeling Tasks with a Probabilistic Language Adaptation Model