Efficient Learning of Sparse Representations from Interactions Conference: Proceedings of the ACM Web Conference 2026; April 13–17, 2026; Dubai, United Arab Emirates Proceedings of the ACM Web Conference 2026 (WWW ’26), April 13–17, 2026, Dubai, United Arab Emirates DOI: 10.1145/3774904.3792914 ISBN
Vojtěch Vančura, Martin Spišák, Rodrigo Alves, Ladislav Peška · 2026
Read Efficient Learning of Sparse Representations from Interactions Conference: Proceedings of the ACM Web Conference 2026; April 13–17, 2026; Dubai, United Arab Emirates Proceedings of the ACM Web Conference 2026 (WWW ’26), April 13–17, 2026, Dubai, United Arab Emirates DOI: 10.1145/3774904.3792914 ISBN by Vojtěch Vančura, Martin Spišák, Rodrigo Alves, Ladislav Peška (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/arxiv-2602-09935.txt.
Description
Recommender systems first narrow millions of items to a short list, and that retrieval step must be both sharp and fast. A 2026 paper led by Vojtěch Vančura, with three co-authors, tackles the tension between expressive embeddings and serving cost. The team swaps dense embedding layers for high-dimensional sparse ones, training them to stay compact without losing expressive power.
Applied to a production-grade collaborative filtering autoencoder called ELSA, the approach shrank embeddings tenfold with no accuracy loss, and a hundredfold with only a small loss. The active dimensions also form an interpretable inverted index that groups items along the model's latent space. That structure lets segment-level features, such as two-dimensional homepage layouts, live inside the retrieval model itself.
Catalog record
| CREATOR | Vojtěch Vančura, Martin Spišák, Rodrigo Alves, Ladislav Peška |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 20 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-09935