Reproduce “Online Algorithms Cannot Find Isolated Perceptron Solutions” 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
Shuangping Li, Gabe Schoenbach
The authors prove that, in the random binary perceptron, every online algorithm has probability o(1) of outputting an h-isolated solution for constraint density α > 0 and margin κ, when h grows faster than log^4 n. Online algorithms commit to each sign before seeing later columns. The result concerns isolated solutions, not feasibility, and allows unbounded computation and memory.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.