Reproduce “Active Learning of Conditional Generative Models via the Transport Neural Tangent Kernel” 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
Jayoung Ryu, Kyunghyun Cho, Romain Lopez
The authors introduce a transport neural tangent kernel (tNTK) for selecting conditions when experiments produce distributions rather than scalar outcomes. It compares parameter sensitivities of a generative transport map across source samples and guides greedy selection through Gaussian-process posterior variance. Across benchmarks, it improves distributional prediction in many settings, including a biological task; evidence concerns predictive performance under fixed acquisition budgets.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.