CANDI - Hybrid Discrete-Continuous Diffusion Models

Learn about hybrid discrete-continuous diffusion models through Patrick Pynadath's 52-minute research presentation from the Generative Memory Lab. Explore the CANDI framework that combines discrete and continuous diffusion processes, examining how this hybrid approach addresses challenges in generative modeling across different data types. Discover the theoretical foundations, implementation details, and practical applications of this novel diffusion model architecture that bridges the gap between discrete and continuous domains in machine learning.