TorchKM: A GPU-Oriented Library for Kernel Learning and Model Selection

About

TorchKM is an open-source library for kernel machines, including support vector machines, kernel logistic regression, and kernel quantile regression, with GPU acceleration. The library features a scikit-learn-style API and is designed to exploit GPU-friendly linear algebra, accelerating the full training and model-selection pipeline through intelligent reuse of matrix operations. Benchmarks show competitive predictive performance with substantial speedups over standard baselines. The efficiency and programmable design also make TorchKM a kernel-learning component for AI-driven workflows. Code and documentation are available at https://github.com/YikaiZhang95/torchkm, and the package can be easily installed via PyPI.

Yikai Zhang, Gaoxiang Jia, Jie Ding, Boxiang Wang · 2026

Related benchmarks

Task	Dataset	Result
Kernel Learning	Simulated Data (n=10000, p=10)	Objective Value0.47	3
Kernel Learning	Simulated Data (n=10000, p=1000)	Objective Value0.48	3
Kernel Learning	Simulated Data (n=20000, p=10)	Objective Value0.45	3
Kernel Learning	Simulated Data (n=20000, p=100)	Objective Value0.28	3
Kernel Learning	Simulated Data n=20000, p=1000	Time (s)129.3	3
Kernel Learning	Simulated Data (n=10000, p=100)	Objective Value0.37	3
Classification	a9a LIBSVM (test)	Accuracy85.1	2
Classification	w8a LIBSVM (test)	Accuracy98.1	2
Classification	ijcnn1 LIBSVM (test)	Accuracy98.2	2
Classification	covtype LIBSVM (test)	Accuracy80.7	2