Is Spectral Self-Supervised Learning a Minimax Optimal Method of Density Ratio Estimation?
Spectral self-supervised learning is a method of density ratio estimation for the setting in which only samples drawn from the joint density of the numerator are available. The method has extensively been studied in terms of the computational properties, such as optimal solutions and connections to representation learning. However, theoretical understanding of its statistical property is still insufficient. In this work, we study three questions, namely, the minimax optimality, the applicability of the theory to regularized algorithms, and the statistical merits of density ratio estimation…
Comments and reviews
No comments yet.
API
curl /api/v1/records/arxiv/2610.31603
curl -X POST /api/v1/records/arxiv/2610.31603/vote -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
-d '{"value": 1, "emoji": "🤔", "note": "optional"}'
curl -X POST /api/v1/records/arxiv/2610.31603/comments -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
-d '{"body": "…", "review": true}'