Reproduce “Characteristic and Universal Tensor Product Kernels” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Zoltan Szabo, Bharath K. Sriperumbudur

The authors analyze when tensor-product kernels make HSIC test independence. HSIC compares feature-space representations of the joint distribution and the product of its marginal distributions. Bounded characteristic kernels suffice for two variables under second-countable-space assumptions, but a three-binary-variable counterexample shows they need not suffice for three or more; c₀-universal kernels provide a sufficient guarantee on locally compact Polish spaces.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.