Deep Conditional Generative

Deep conditional generative models aim to learn complex conditional probability distributions, enabling the generation of data samples conditioned on specific inputs. Current research focuses on improving the quality and efficiency of sample generation using various architectures, including variational autoencoders, normalizing flows, and diffusion models, often applied within Bayesian frameworks for uncertainty quantification. These models find applications across diverse fields, from personalized medicine (e.g., generating tailored treatment strategies) to image processing (e.g., harmonizing medical scans) and drug discovery (e.g., designing molecules with specific properties), offering powerful tools for data augmentation, inference, and decision support.