Uncertainty-Calibrated Collaborative Filtering via Heteroscedastic Bayesian Neural Networks
Xiaowei Wang · 2026
Read Uncertainty-Calibrated Collaborative Filtering via Heteroscedastic Bayesian Neural Networks by Xiaowei Wang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13605793.txt.
Description
Wang presents a 2026 paper on recommendation systems that know how unsure they are. Neural recommenders give point ratings with no sense of confidence. Past fixes capture only doubt from limited data, missing the noise inside the ratings themselves. The new model, BayesNCF-v2, joins a Bayesian output layer with a head that learns per-case noise.
Across three data sets and five runs, confidence errors fell by a quarter to near three-quarters versus the plain model. Scores matched far heavier ensemble systems with far fewer parameters. For brand-new users, the model widened its doubt on its own, which tamed its errors. Both random and time-based test splits backed the result.
Catalog record
| CREATOR | Xiaowei Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 39 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |