Training Data Protection with Compositional Diffusion Models

arXiv:2308.01937v3 Announce Type: replace-cross
Abstract: We introduce Compartmentalized Diffusion Models (CDM), a method to train different diffusion models (or prompts) on distinct data sources and arbitrarily compose them at inference time. The individual models can be trained in isolation, at different times, and on different distributions and domains and can be later composed to achieve performance comparable to a paragon model trained on all data simultaneously. Furthermore, each model only contains information about the subset of the data it was …

arxiv can cdm compose cs.ai cs.cr cs.cv cs.lg data data protection data sources diffusion models distributions domains isolation prompts protection train training training data