Support Vector Machines Directly Minimise KL Divergence
Abstract
Support vector machines (SVMs) are fundamentally grounded in geometric margin maximisation and currently lack a mainstream probabilistic interpretation. This keeps them theoretically isolated from probabilistic machine learning models that are naturally comparable and combinable. In this paper, we bridge this gap by proving that the dual objectives of both support vector classification and support vector regression are exact, constrained Kullback-Leibler (KL) divergence minimisation tasks between two Gaussian measures. We further show that this same divergence formulation recovers the exact objective function for kernel ridge regression when domain constraints are removed and L2 regularisation is applied. By demonstrating that SVM duals directly minimise relative entropy, our results provide a unified information-theoretic model for standard kernel machines and lay the foundation for unifying broader kernel-based algorithms under probability theory.
Related
- Support Vector Machines for Classification, Regression, and Outlier Detection: A Tutorial
- Comparative Performance Analysis of Optimization Solvers for Linear Support Vector Machines: Impact of Preconditioning and Dimensionality An Empirical Study Using the Benchopt Framework
- High-Dimensional Statistical Inference for Sparse Support Vector Machines
- Representative Learning for Distributed Support Vector Machines
- Asymptotic Properties of Support Vector Machines in High-Dimension, Low-Sample-Size Settings under a Spiked Model
BibTeX
@misc{silajev2026support,
title = {{Support Vector Machines Directly Minimise KL Divergence}},
author = {Ivan Silajev},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.05557},
eprint = {2611.05557},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.05557},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - Support Vector Machines Directly Minimise KL Divergence
AU - Ivan Silajev
PY - 2026
DA - 2026/11/03
PB - arXiv
DO - 10.48550/arxiv.2611.05557
UR - https://arxiv.org/abs/2611.05557
AB - Support vector machines (SVMs) are fundamentally grounded in geometric margin maximisation and currently lack a mainstream probabilistic interpretation. This keeps them theoretically isolated from probabilistic machine learning models that are naturally comparable and combinable. In this paper, we bridge this gap by proving that the dual objectives of both support vector classification and support vector regression are exact, constrained Kullback-Leibler (KL) divergence minimisation tasks between two Gaussian measures. We further show that this same divergence formulation recovers the exact objective function for kernel ridge regression when domain constraints are removed and L^2 regularisation is applied. By demonstrating that SVM duals directly minimise relative entropy, our results provide a unified information-theoretic model for standard kernel machines and lay the foundation for unifying broader kernel-based algorithms under probability theory.
ER -