Non-convex SGD and Lojasiewicz-type Conditions for Deep Learning

Explore a conference talk on non-convex stochastic gradient descent (SGD) and Lojasiewicz-type conditions for deep learning, presented by Kevin Scaman at the Centre International de Rencontres Mathématiques in Marseille, France. Delve into advanced mathematical concepts applied to machine learning optimization techniques during this 47-minute presentation, recorded as part of the "Learning and Optimization in Luminy" thematic meeting. Access this talk and other presentations by renowned mathematicians through CIRM's Audiovisual Mathematics Library, featuring chapter markers, keywords, enriched content with abstracts and bibliographies, and a multi-criteria search function for easy navigation and in-depth exploration of mathematical topics.