Predictive Coding: A Biologically Plausible Alternative to Backpropagation

This 27-minute video explores Predictive Coding as a biologically plausible alternative to the backpropagation algorithm, deriving it from first principles. Discover how neural networks might learn in biological systems through a comprehensive explanation of the credit assignment problem, limitations of backpropagation in biological contexts, and the theoretical foundations of predictive coding. Follow along as the presenter, a graduate student at NYU Center for Neural Science and researcher at Flatiron Institute, explains energy formalism, activity update rules, neural connectivity patterns, and weight update mechanisms that make predictive coding work. The presentation systematically builds up the mathematical framework before demonstrating how all components integrate into a complete learning algorithm for biological networks. The video includes detailed references to scientific literature and follows a structured outline covering introduction, theoretical foundations, mathematical derivations, and practical implementations of predictive coding in neural systems.