Reproduce “An Entropic Associative Memory” 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
Luis A. Pineda, Gibrán Fuentes, Rafael Morales
The paper presents an associative memory that recognizes and reconstructs MNIST digits from feature cues rather than storage addresses. An encoder represents digits with 64 discrete features; registers store possible values, test containment, and construct outputs from allowed values. In experiments, precision declined slightly while recall rose as registers filled, and retrieval quality began declining around entropy 3.0 for some digits.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.