Learning Real-World Probabilistic Models with Approximate Message Passing

Explore a comprehensive lecture on learning real-world probabilistic models using approximate message passing. Delve into the challenges posed by big and structured data in statistical inference and decision-making. Examine the shifting assumptions in data modeling, including parameter storage, granularity of building blocks, and the interplay between computation, storage, communication, and inference techniques. Discover factor graphs as a versatile modeling technique that combines systems and statistical properties. Review distributed message passing and other approximate inference techniques. Gain insights into real-world applications at Amazon and understand the implications of big data for Statistics and the convergence of statistical models and distributed systems.

Ralf Herbrich: "Learning Real-World Probabilistic Models with Approximate Message Passing"