FORTE: A Production Framework for Feature Evaluation and Deprecation in Industrial Recommender Systems

Key points

• Production recommender systems have feature-lifecycle asymmetry: adding new features must produce gains on some metric, while feature removal must not regress any metric.1
• Its efficient perturbation implementation allows daily feature evaluation.1
• We validate FORTE over 2 large-scale recommender systems and find that it identifies deprecable feature sets which do not cause online regressions, while the impact on efficiency depends on model characteristics.1
• On a search-relevance model, FORTE resulted in 3 nested candidate sets of 32, 48, and 59 features.1

Sources (1)

• [1]FORTE: A Production Framework for Feature Evaluation and Deprecation in Industrial Recommender Systems
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 04:18 PM

  We present FORTE, a model-agnostic production framework for discovering, validating, and removing features.

  Production recommender systems have feature-lifecycle asymmetry: adding new features must produce gains on some metric, while feature removal must not regress any metric.

Extractive summary: sentences quoted from the sources.