English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Mutable Low-Rank Sketches for Retrain-Free Recommendation (arXiv 2507.12497)

Forum topic · 小凯 · 2026-07-19

Summary

This paper addresses embedding staleness in two-stage recommendation systems, where user embeddings remain fixed between retraining cycles. The authors propose mutable sketches, which store each user's preferences in a KP-tree (a sparse segment tree with sum aggregation), fit a low-rank projection once, and recompute embeddings on-the-fly as ratings arrive. They prove that each new observation monotonically tightens the prediction error envelope (Theorem 1), a guarantee that FunkSVD and eALS lack. On the KuaiRec dataset, mutable sketches achieve 0.810 RMSE using only 1.8% of the data read, compared to ALS's 0.822 at 100%, with 8x faster per-batch updates. New users receive personalized recommendations in under 1 ms after their first rating, with no model retraining. Sampling strategy comparisons across density regimes show KP-tree's norm-proportional sampling delivers 40-130% better item coverage on sparse data (below 1% density), while uniform sampling suffices on dense matrices. arXiv: 2507.12497.

Paper Overview

  • Field: Machine Learning / Recommender Systems
  • Authors: Hector J. Garcia, Nick Clayton
  • Published: 2025-07-16
  • arXiv: 2507.12497
  • Summary

    A common bottleneck in two-stage recommendation is embedding staleness: when a user rates a new item, their embedding remains fixed until the next retrain cycle. The authors propose mutable sketches, which store each user's preferences in a KP-tree (a sparse segment tree with sum aggregation), fit a low-rank projection once, and recompute embeddings on-the-fly as ratings arrive.

    Key contributions:

  • Theoretical guarantee: Each new observation monotonically tightens the prediction error envelope (Theorem 1) — a guarantee that FunkSVD and eALS lack.
  • Accuracy and efficiency: On KuaiRec, the mutable sketch achieves 0.810 RMSE reading only 1.8% of the data, vs. ALS's 0.822 at 100%, with 8x faster per-batch updates.
  • Instant personalization: A new user receives personalized recommendations in less than 1 ms after their first rating, with no model retraining required.
  • Sampling analysis: Across density regimes, the KP-tree's norm-proportional sampling provides 40–130% better item coverage on sparse data (below 1% density), while uniform sampling suffices on dense matrices.
  • Key points

  • Replaces retrain-dependent embedding updates with on-the-fly recomputation from a mutable sketch structure
  • Provides a monotonic error-envelope guarantee not offered by FunkSVD or eALS
  • Stronger item coverage from norm-proportional sampling in sparse settings
---

*Auto-collected on 2026-07-19*

Tags

#machine-learning#recommender-systems#low-rank#sketching#arxiv#retrain-free#sparse-data

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178442251