Combining simple elements from the literature, we define a linear model that is geared toward sparse data, in particular implicit feedback data for recommender systems. We show that its training objective has a closed-form solution, and discuss the resulting conceptual insights. Surprisingly, this simple model achieves better ranking accuracy than various state-of-the-art collaborative-filtering approaches, including deep non-linear models, on most of the publicly available data-sets used in our experiments.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Recommendation Systems | MovieLens 20M | Recall@20 | 0.391 | EASE |
| Recommendation Systems | MovieLens 20M | Recall@50 | 0.521 | EASE |
| Recommendation Systems | MovieLens 20M | nDCG@100 | 0.42 | EASE |
| Recommendation Systems | Million Song Dataset | Recall@20 | 0.333 | EASE |
| Recommendation Systems | Million Song Dataset | Recall@50 | 0.428 | EASE |
| Recommendation Systems | Million Song Dataset | nDCG@100 | 0.389 | EASE |
| Recommendation Systems | Netflix | Recall@20 | 0.362 | EASE |
| Recommendation Systems | Netflix | Recall@50 | 0.445 | EASE |
| Recommendation Systems | Netflix | nDCG@100 | 0.393 | EASE |