Generative Bayesian Hyperparameter Tuning
By: Hedibert Lopes, Nick Polson, Vadim Sokolov
\noindent Hyper-parameter selection is a central practical problem in modern machine learning, governing regularization strength, model capacity, and robustness choices. Cross-validation is often computationally prohibitive at scale, while fully Bayesian hyper-parameter learning can be difficult due to the cost of posterior sampling. We develop a generative perspective on hyper-parameter tuning that combines two ideas: (i) optimization-based approximations to Bayesian posteriors via randomized, weighted objectives (weighted Bayesian bootstrap), and (ii) amortization of repeated optimization across many hyper-parameter settings by learning a transport map from hyper-parameters (including random weights) to the corresponding optimizer. This yields a ``generator look-up table'' for estimators, enabling rapid evaluation over grids or continuous ranges of hyper-parameters and supporting both predictive tuning objectives and approximate Bayesian uncertainty quantification. We connect this viewpoint to weighted $M$-estimation, envelope/auxiliary-variable representations that reduce non-quadratic losses to weighted least squares, and recent generative samplers for weighted $M$-estimators.
Similar Papers
Bayesian Optimization under Uncertainty for Training a Scale Parameter in Stochastic Models
Machine Learning (CS)
Finds best settings for tricky computer programs faster.
From Black-Box Tuning to Guided Optimization via Hyperparameters Interaction Analysis
Machine Learning (CS)
Helps computers learn better by finding best settings.
High-Dimensional Surrogate Modeling for Closed-Loop Learning of Neural-Network-Parameterized Model Predictive Control
Machine Learning (CS)
Teaches robots to learn better with more settings.