Fast Inference for Probabilistic Graphical Models
Offered By: USENIX via YouTube
Course Description
Overview
Explore a conference talk on Fast-PGM, an innovative system for fast and parallel inference in probabilistic graphical models (PGMs). Delve into the challenges faced by existing PGM inference systems, including inefficiency and lack of generality. Learn about Fast-PGM's approach to overcoming these issues through careful memory management techniques, computation and parallelization optimizations, and a flexible system design. Discover how Fast-PGM achieves significant speedups over state-of-the-art implementations while maintaining high generality and ease of integration with mainstream importance sampling-based algorithms. Gain insights into the system's architecture, which facilitates easy optimizations, extensions, and customization for users. Understand the potential impact of Fast-PGM on improving the efficiency and applicability of PGMs in various domains.
Syllabus
USENIX ATC '24 - Fast Inference for Probabilistic Graphical Models
Taught by
USENIX
Related Courses
Probabilistic Graphical Models 1: RepresentationStanford University via Coursera Probabilistic Graphical Models 2: Inference
Stanford University via Coursera Probabilistic Graphical Models 3: Learning
Stanford University via Coursera Artificial Intelligence
Udacity Probabilistic Graphical Models
Stanford University via Coursera