The Rise of Tractable Circuits - From Cryptography to Continuous Generative Models

Explore the advancements in tractable circuits through this 44-minute lecture by Robert Peharz from TU Graz. Delve into the intersection of neural networks, logic, and probabilistic models as Peharz demonstrates how tractable circuits contribute to reliable AI reasoning. Discover their application in side-channel attacks on crypto systems, combining logical knowledge, learning, and probabilistic inference for fast and nearly exact inference. Examine the extension of probabilistic circuits to continuous latent variable models, introducing a new paradigm in circuit learning and inference that achieves state-of-the-art results in various applications.