Richard E. Turner

Papers

Total Citations

H-Index

About

Richard E. Turner is a leading researcher in probabilistic machine learning, with a particular focus on advancing inference and optimization techniques for complex, high-dimensional problems. His work spans Bayesian deep learning, scalable variational inference, and probabilistic numerics, where he develops principled methods that balance theoretical rigor with practical performance. Turner’s major contributions include pioneering the use of structured random orthogonal matrices for gradient approximation in blackbox optimization, as demonstrated in his highly cited work on structured evolution for scalable policy optimization (52 citations). This approach provides provably more accurate estimators than traditional baselines, enabling superior learning in reinforcement learning and optimization tasks. Additionally, Turner has made significant strides in modeling circular data, a domain often neglected in machine learning, through his development of the multivariate generalised von Mises distribution. His work on this topic (11 citations) extends standard probabilistic tools to circular variables, with applications ranging from robotics to social sciences. Turner’s research is characterized by its impact on both theory and practice, making him a key figure in advancing probabilistic methods for modern AI challenges.