Coursera Spring Sale
40% Off Coursera Plus Annual!
Grab it
This seminar explores the phenomenological aspects of probabilistic modeling as a form of perception with specific abilities and limitations. Delve into how mathematical pragmatics of statistical models function as a "body schema" of perceptual ability, particularly focusing on undirected models and their parametrization. Examine the challenges these models present for exact and approximate inference through the lens of image segmentation and object recognition using Markov networks and conditional random fields. Learn how concepts like conditioning, dependencies, cliques, and factor reduction operate within these frameworks. The presentation connects probability to experience within mathematical modeling practices themselves, drawing on Husserl and Merleau-Ponty's ideas about movement being essential to cognition. Explore how statistical conjugacy requires both formal necessity and "slippery motility," with category theory applied to understand the inherent structures of these models. The fundamental question addressed is "what can inference do?" and what this reveals about its overall schema of perception.
