Reproduce “Revisiting the Performance of iALS on Item Recommendation Benchmarks” 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

Steffen Rendle, Walid Krichene, Li Zhang, Yehuda Koren

The paper revisits implicit alternating least squares (iALS) for item recommendation, tuning embedding size, unobserved-pair weighting, and frequency-scaled regularization on four established benchmarks. iALS represents users and items as vectors and alternates least-squares updates. On the Million Song Dataset, its Recall@20 is 0.309, versus 0.211 previously, but 0.333 for EASE; results are benchmark-specific, not universal wins.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.