Boltzmann Tuning of Generative Models

TL;DR

This paper introduces Boltzmann Tuning of Generative Models (BTGM), a versatile method for optimizing generative models to produce high-quality instances based on external criteria, offering an efficient alternative to rejection sampling.

Contribution

It formalizes the tuning as a well-posed optimization problem and provides a practical methodology for selecting the best criteria for model tuning.

Findings

Demonstrates BTGM's effectiveness in a real-world energy policy design task.

Shows BTGM can sample extreme regions of the criteria effectively.

Offers an affordable alternative to rejection sampling for model tuning.

Abstract

The paper focuses on the a posteriori tuning of a generative model in order to favor the generation of good instances in the sense of some external differentiable criterion. The proposed approach, called Boltzmann Tuning of Generative Models (BTGM), applies to a wide range of applications. It covers conditional generative modelling as a particular case, and offers an affordable alternative to rejection sampling. The contribution of the paper is twofold. Firstly, the objective is formalized and tackled as a well-posed optimization problem; a practical methodology is proposed to choose among the candidate criteria representing the same goal, the one best suited to efficiently learn a tuned generative model. Secondly, the merits of the approach are demonstrated on a real-world application, in the context of robust design for energy policies, showing the ability of BTGM to sample the extreme regions of the considered criteria.