Beren Millidge

Papers

Total Citations

H-Index

About

Beren Millidge is a computational neuroscience and machine learning researcher whose work bridges theoretical neuroscience and artificial intelligence, with a particular focus on biologically plausible learning algorithms and probabilistic frameworks for cognition. He is best known for his contributions to **active inference** and **predictive coding** — two deeply interconnected theories about how the brain implements perception, action, and learning through generative models and free energy minimization. His highly cited 2021 survey on active inference in robotics and artificial agents (55 citations) has become an essential reference for researchers exploring this framework beyond neuroscience, demonstrating its promise for state-estimation and control under uncertainty in artificial systems. Complementing this, his work on predictive coding as an alternative to backpropagation has attracted significant attention, positioning it as a biologically grounded and potentially more parallelizable approach to training deep neural networks — an increasingly urgent question as the field grapples with the limitations of conventional deep learning. Millidge's research is notable for making mathematically demanding frameworks accessible and practically relevant, helping translate theoretical neuroscience into actionable directions for AI development. His work appeals broadly to researchers interested in the future of brain-inspired computing and next-generation learning algorithms.