Reproduce “Learning from Aggregate responses: Instance Level versus Bag Level Loss Functions” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Adel Javanmard, Lin Chen, Vahab Mirrokni, Ashwinkumar Badanidiyuru, Gang Fu
The authors study individual prediction from group-average responses, introducing a loss that blends fitting each average with restraining prediction differences within a group. For squared loss, instance-level fitting equals group-level fitting plus a dispersion penalty. In Gaussian theory, the lower-risk endpoint depends on signal-to-noise ratio; on Boston Housing, optimal ρ was 0 for bag size 80 and 1 for 240.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.