Generative Bayesian Hyperparameter Tuning

TL;DR

This paper introduces a novel generative approach to hyperparameter tuning that combines approximate Bayesian posteriors with amortized optimization, enabling fast evaluation and uncertainty quantification across hyper-parameter ranges.

Contribution

It proposes a generative framework that integrates weighted Bayesian bootstrap and amortized optimization to improve hyper-parameter tuning efficiency and Bayesian uncertainty estimation.

Findings

Enables rapid evaluation of hyper-parameters using a generator lookup table.

Supports predictive tuning objectives and Bayesian uncertainty quantification.

Connects hyper-parameter tuning to weighted M-estimation and recent generative samplers.

Abstract

\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.