Core-Elements Subsampling for Alternating Least Squares

A 110.7 MB research artifact by Dunyao Xue, last updated in June 2026, proposes a novel subsampling method for the Alternating Least Squares algorithm. The work focuses on low-rank matrix factorization for datasets with missing values, a common challenge in building recommender systems. The included files demonstrate the method's theoretical guarantees and its application in simulations and real-world scenarios.

Use Cases

• Benchmarking computational efficiency improvements for ALS-based recommender systems based on the described subsampling method.
• Studying approximation techniques for matrix factorization with missing data as outlined in the paper.
• Reproducing simulation results for core-elements selection in large-scale user-item interaction data.

Strengths

• Includes theoretical guarantees for approximation and convergence, as stated in the description.
• Demonstrated with extensive simulations and real-world applications, according to the author.
• Released under a permissive CC-BY-4.0 license, facilitating reuse and modification.

Limitations

• The actual dataset rows and columns are unknown; data quality and structure require inspection after download.
• Column-level documentation is absent; field semantics must be inferred from the accompanying research paper.
• The primary content appears to be a research paper (PDF) and code/artifacts (ZIP), not a standalone, directly usable dataset.

Provenance

Source

Dunyao Xue

Collection Method

Proposed as a novel computational method; data likely generated for simulations and used in example applications.

Freshness

Last updated 2026-06-03 10:54:02