2025/03/20 by Alexandre Vérine, Alexandre Verine, Verine, Alexandre +9 · 1 voice
Computer Science · #Face and Expression Recognition #Face recognition and analysis #Generative Adversarial Networks and Image Synthesis #cs.LG
paper · pdf · doi:10.48550/arxiv.2503.16117
arxiv published 2025/03/20 · arxiv updated 2025/06/11
Discriminator Guidance has become a popular method for efficiently refining pre-trained Score-Matching Diffusion models. However, in this paper, we demonstrate that the standard implementation of this technique does not necessarily lead to a distribution closer to the real data distribution. Specifically, we show that training the discriminator using Cross-Entropy loss, as commonly done, can in fact increase the Kullback-Leibler divergence between the model and target distributions, particularly when the discriminator overfits. To address this, we propose a theoretically sound training objective for discriminator guidance that properly minimizes the KL divergence. We analyze its properties and demonstrate empirically across multiple datasets that our proposed method consistently improves over the conventional method by producing samples of higher quality.