Papers › Why is Normalization Necessary for Linear Recommenders?

Why is Normalization Necessary for Linear Recommenders?

8 Apr 2025arXiv:2504.05805archive 2025-07-28

Seongmin Park, Mincheol Yoon, Hye-Young Kim, Jongwuk Lee

Despite their simplicity, linear autoencoder (LAE)-based models have shown comparable or even better performance with faster inference speed than neural recommender models. However, LAEs face two critical challenges: (i) popularity bias, which tends to recommend popular items, and (ii) neighborhood bias, which overly focuses on capturing local item correlations. To address these issues, this paper first analyzes the effect of two existing normalization methods for LAEs, i.e., random-walk and symmetric normalization. Our theoretical analysis reveals that normalization highly affects the degree of popularity and neighborhood biases among items. Inspired by this analysis, we propose a versatile normalization solution, called Data-Adaptive Normalization (DAN), which flexibly controls the popularity and neighborhood biases by adjusting item- and user-side normalization to align with unique dataset characteristics. Owing to its model-agnostic property, DAN can be easily applied to various LAE-based models. Experimental results show that DAN-equipped LAEs consistently improve existing LAE-based models across six benchmark datasets, with significant gains of up to 128.57% and 12.36% for long-tail items and unbiased evaluations, respectively. Refer to our code in https://github.com/psm1206/DAN.

PaperPDFCode

Code

psm1206/dan officialmentioned in papermentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Collaborative FilteringRecommendation Systems

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Collaborative Filtering Amazon-Book RLAE-DAN NDCG@20 0.0630 #2 of 6 Archive leaderboard report
Collaborative Filtering Amazon-Book RLAE-DAN Recall@20 0.0762 #2 of 6 Archive leaderboard report
Collaborative Filtering Gowalla RLAE-DAN NDCG@20 0.1605 #1 of 11 Archive leaderboard report
Collaborative Filtering Gowalla RLAE-DAN Recall@20 0.1922 #1 of 11 Archive leaderboard report
Collaborative Filtering Yelp2018 RLAE-DAN NDCG@20 0.0587 #3 of 9 Archive leaderboard report
Collaborative Filtering Yelp2018 RLAE-DAN Recall@20 0.0706 #3 of 9 Archive leaderboard report
Recommendation Systems Amazon-Book RLAE-DAN Recall@20 0.0762 #3 of 16 Archive leaderboard report
Recommendation Systems Amazon-Book RLAE-DAN nDCG@20 0.0630 #3 of 16 Archive leaderboard report
Recommendation Systems Gowalla RLAE-DAN Recall@20 0.1922 #3 of 13 Archive leaderboard report
Recommendation Systems Gowalla RLAE-DAN nDCG@20 0.1605 #3 of 13 Archive leaderboard report
Recommendation Systems Yelp2018 RLAE-DAN NDCG@20 0.0587 #3 of 11 Archive leaderboard report
Recommendation Systems Yelp2018 RLAE-DAN Recall@20 0.0706 #3 of 11 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

ALIGNSPEED

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections