We study the statistical and computational aspects of kernel principal component analysis using random Fourier features and show that under mild assumptions, O(nlogn) features suffices to achieve O(1/ϵ2) sample complexity. Furthermore, we give a memory efficient streaming algorithm based on classical Oja's algorithm that achieves this rate.
Cite
@article{arxiv.1808.00934,
title = {Streaming Kernel PCA with $\tilde{O}(\sqrt{n})$ Random Features},
author = {Enayat Ullah and Poorya Mianjy and Teodor V. Marinov and Raman Arora},
journal= {arXiv preprint arXiv:1808.00934},
year = {2018}
}
Comments
Advances in Neural Information Processing Systems (NIPS), 2018. 42 pages, 3 figures