Reproduce “Function Encoders: A Principled Approach to Transfer Learning in Hilbert Spaces” 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
Tyler Ingebrand, Adam J. Thorpe, Ufuk Topcu
The authors introduce a function encoder that adapts to a new task from a few examples by combining learned neural-network basis functions. Its least-squares coefficients do not require the bases to be orthonormal. In selected experiments, it achieves low squared L2 error on polynomial tasks, including type-2 and type-3 transfer, but this does not establish transfer to arbitrary unseen functions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.