Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
With the growing amount of information available online, recommender systems are starting to provide a viable alternative and complement to search engines, in helping users to find objects of interest. Methods based on Matrix Factorization (MF) models are the state-of-the-art in recommender systems. The input to MF is user feedback, in the form of a rating matrix. However, users can be engaged in interactions with multiple types of entities across different contexts, leading to multiple rating matrices. In other words, users can have interactions in a heterogeneous information network. Generally, in a heterogeneous network, entities from any two entity types can have interactions with a weight (rating) indicating the level of endorsement. Collective Matrix Factorization (CMF) has been proposed to address the recommendation problem in heterogeneous networks. However, a main issue with CMF is that entities share the same latent factor across different contexts. This is particularly problematic in two cases: Latent factors for entities that are cold-start in a context will be learnt mainly based on the data from other contexts where these entities are not cold-start, and therefore the factors are not properly learned for the cold-start context. Also, if a context has more data compared to another context, then the dominant context will dominate the learning process for the latent factors for entities shared in these two contexts. In this paper, we propose a context-dependent matrix factorization model, HeteroMF, that considers a general latent factor for entities of every entity type and context-dependent latent factors for every context in which the entities are involved. We learn a general latent factor for every entity and transfer matrices for every context to convert the general latent factors into a context-dependent latent factor. Experiments on two real life datasets from Epinions and Flixster demonstrate that HeteroMF substantially outperforms CMF, particularly for cold-start entities and for contexts where interactions in one contexts are dominated by other contexts.
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.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.001 |
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 it