Reproduce “Principled Non-Linear Feature Selection” 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
Dimitrios Athanasakis, John Shawe-Taylor, Delmiro Fernandez-Reyes
randSel selects features for nonlinear prediction by estimating each feature’s contribution to kernel-based dependence between features and labels across matched random subsets, then repeatedly removing low contributors. In a 200-feature XOR illustration, the two planted features become more distinct after culling. On signal-peptide prediction, randSel reports 72.75% accuracy versus RFE’s 71.81%; adding MKL reaches 75.28%, without establishing statistical significance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.