Get Started
Home
Topics
Search
Library
Research questionCan one squared-loss estimator achieve both minimax and universal exponential rates for finite versus countably infinite hypothesis classes?Model-selection aggregation seeks minimax excess-risk guarantees, while universal learning seeks exponential rates. Whether one estimator can provide both depends on whether the hypothesis class is finite or countably infinite.
AI
Machine Learning
Research Paper
Statistical Machine Learning
Latest papersRecent research connected to this question, newest first.Reconciling Universal and Uniform Learning with $Q$-AggregationThe evidence covers finite and countably infinite hypothesis classes under squared loss with bounded responses. For finite classes, Q-aggregation achieves both guarantees, whereas for countably infinite classes the source identifies an inherent trade-off and characterizes it through combining algorithms optimized for each goal.research paper · Sep 4, 2026
Related questions
How can Bayes-optimal BER and AUC be estimated from corrupted soft labels with unknown class priors?What convergence guarantees and irreducible errors arise in random-reshuffling optimization for finite sums and composite objectives?How can noisy Lipschitz regression recover an unknown function uniformly on a general metric space?How can agnostic learning handle low-intrinsic-dimensional concepts under arbitrary distributions with Gaussian-robust benchmarks?