Fine-Tuning of Continuous-Time Diffusion Models as Entropy-Regularized Control

arXiv:2402.15194v1 Announce Type: new
Abstract: Diffusion models excel at capturing complex data distributions, such as those of natural images and proteins. While diffusion models are trained to represent the distribution in the training dataset, we often are more concerned with other properties, such as the aesthetic quality of the generated images or the functional properties of generated proteins. Diffusion models can be finetuned in a goal-directed way by maximizing the value of some reward function (e.g., the aesthetic quality of …

abstract arxiv continuous control cs.ai cs.lg data dataset diffusion diffusion models distribution entropy excel fine-tuning generated images natural proteins quality stat.ml training type