The Promises and Pitfalls of Stochastic Gradient Langevin Dynamics - Eric Moulines

Explore the promises and pitfalls of Stochastic Gradient Langevin Dynamics in this 52-minute keynote talk by Eric Moulines at the Alan Turing Institute. Delve into the analysis of adaptive stochastic gradient and MCMC algorithms, examining their applications in data processing and machine learning. Gain insights into the challenges faced in implementing these state-of-the-art algorithms for high-dimensional real-world problems. Learn about the importance of stochastic approximations and MCMC algorithms in Data Science, and understand their adaptive nature in tracking key parameters of data streams with changing dynamics. Discover how leading experts in stochastic approximations, MCMC algorithms, Bayesian inference, and probability theory address technical and theoretical challenges in the field.