Reproduce “Learning Sparse Representations of Multimodal Content for Enhanced Cold Item Recommendation” 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

Gregor Meehan, Johan Pauwels

The authors develop sparse multimodal content representations for recommending items with no interaction history. Their method retains a small number of item-vector dimensions and applies a pre-sparsification activation to sharpen similarities; users’ vectors are pruned. On M4A-Onion, it achieves NDCG@20 of 0.1350 versus 0.0769 for the dense comparison, under a matched nominal storage budget; gains are not uniform across datasets.

Summary by alphaXiv.

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