Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Learn about a novel framework for amortized probabilistic clustering through this machine learning lecture presented by Irit Chelly from Ben-Gurion University. Explore GFNCP (Generative Flow Network Clustering Process), an innovative approach that addresses the order-dependency issues found in existing neural clustering methods like the Neural Clustering Process. Discover how GFNCP uses Generative Flow Networks with shared energy-based parametrization of policy and reward to achieve consistent clustering posterior under marginalization and order invariance. Examine the mathematical foundations showing how flow matching conditions relate to clustering consistency, and review performance comparisons demonstrating GFNCP's superiority over existing methods on both synthetic and real-world datasets. Gain insights into the intersection of probabilistic clustering, non-parametric Bayesian models, and generative models from a researcher specializing in vision, inference, and learning. The presentation is based on research published in AISTATS '25 and includes discussion of practical applications in unsupervised learning scenarios where traditional Markov chain methods are computationally prohibitive.
