Reproduce “Functional Output Regression with Infimal Convolution: Exploring the Huber and $ε$-insensitive Losses” 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
Alex Lambert, Dimitri Bouche, Zoltan Szabo, Florence d'Alché-Buc
The authors extend functional output regression with Huber and epsilon-insensitive losses defined by norms of whole-function residuals, using duality and finite representations to make the optimization problems tractable. Their experiments find Huber variants outperform squared loss under tested global synthetic corruption, while variants differ under local corruption. These results support robustness in tested settings, not universal superiority across corruptions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.