Variational autoencoders with Gaussian mixture prior for \nrecommender systems
Bibliographic record
Abstract
Recommender systems are used everywhere, from search engines to entertainment websites like video or audio streaming platforms. They are essential when the amount of information available is substantial by providing users with the right information at the right time. The standard approach involves gathering the impressions of users based on content to create the next recommendation. However, collecting these impressions is costly and is not always accurate. A different alternative is to simply gather the binary interactions between a user and the content. \n \nRadio-Canada, the official French broadcaster in Canada, has collected these types of interactions for its video streaming platform called “Tou.TV”. They asked us to create a novel recommender system using temporal and contextual signals. In this thesis, we present two hybrid systems based on a variational autoencoder (VAE) architecture with a Gaussian mixture prior to better understand the latent space with multiple Gaussian distributions. We apply these systems on the Tou.TV dataset, but also demonstrate the efficacy of our approach on the popular dataset called MovieLens-20M. These systems with the simple enhancements proposed on the traditional VAE architecture are able to outperform popular off-the-shelf models.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.004 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".