Answers are grounded in this paper and cite their sources.
Figure 1: Overview of SAGEA: The training phase (left) consists of: (a) training the ELSA collaborative autoencoder on individual user interactions to learn dense user embeddings, and (b) training a Top- sparse autoencoder (SAE) to reconstruct these embeddings via a disentangled, sparse representation.The disentangled representation can be understood as a dictionary of activated concepts (such as preference for "Old music", "Love", "Dark humor", etc., as highlighted in the example).During inference (right), individual group members are encoded into sparse representations via the ELSA encoder E extended with the SAE encoder E .These are aggregated using a fairness-preserving strategy to form a group-level sparse embedding.The SAE decoder (D ) transforms this back into a dense embedding, which is then decoded by the ELSA decoder (D ) to generate group-wise relevance scores.
No takes yet. Share an insight, caveat, or question.
Koštejn et al. (2025) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: