Kenji Doya

Kenji Doya is a Japanese neuroscientist and computational neuroscientist known for contributions to models of basal ganglia, reinforcement learning, and neural implementations of decision-making. He has held research positions at institutions including ATR (advanced telecommunications research institute international), RIKEN, and Kyoto University, and has influenced work in artificial intelligence and robotics through interdisciplinary collaborations. Doya's research bridges experimental neurophysiology, theoretical modeling, and machine learning applications.

Early life and education

Doya was born in Tokyo and completed undergraduate and graduate studies at the University of Tokyo where he trained in biosciences and biophysics alongside mentors connected to laboratories such as those of Toshio Yanagida and investigators affiliated with NTT research groups. During his doctoral period he engaged with experimental groups studying molecular motors and neuronal dynamics, interacting with researchers from institutions like Kyoto University and collaborating indirectly with scientists associated with Osaka University and Tohoku University. His early exposure to labs and conferences involving figures from Cold Spring Harbor Laboratory and meetings of the Society for Neuroscience influenced his transition toward computational approaches.

Academic career

Doya's postdoctoral and faculty career includes appointments at industrial and governmental research centers such as ATR and RIKEN where he led teams integrating computational modeling with electrophysiology and behavioral experiments. He held professorial and visiting positions at universities including Kyoto University, served on advisory panels for the Japan Society for the Promotion of Science, and participated in international collaborations with groups at institutions like MIT, Stanford University, and the Max Planck Society. Doya organized symposia at conferences including the Neural Information Processing Systems and the International Joint Conference on Artificial Intelligence and mentored students who later joined labs at University College London, University of California, Berkeley, and ETH Zurich.

Research contributions

Doya pioneered computational models of the basal ganglia focusing on how neuromodulators implement reinforcement learning algorithms in the brain, connecting theories from Richard Sutton and Andrew Barto to physiology described in studies by Anne Graybiel and W. E. Skaggs. He proposed distinctions among dopaminergic, serotonergic, noradrenergic, and cholinergic systems in terms of learning rules and control, relating to experimental findings from labs of Wolfram Schultz and Read Montague. Doya developed biologically plausible algorithms for actor-critic architectures, linking to work in temporal-difference learning and contributions by Christopher Watkins and Peter Dayan. He advanced models of synaptic plasticity informed by spike-timing research from groups led by Henry Markram and Gerald Edelman, and incorporated notions of stochastic optimal control related to path integral control used in robotics labs at University of Pennsylvania and Stanford University. His interdisciplinary work influenced developments in deep reinforcement learning alongside researchers from DeepMind and academic groups at Carnegie Mellon University and impacted implementations in humanoid and mobile robotics developed at Honda, Toyota, and academic robotics labs such as Janelia Research Campus-affiliated teams. Doya's theoretical proposals on exploration-exploitation trade-offs, neuromodulatory roles, and hierarchical control intersect with studies from Yoshua Bengio, Geoffrey Hinton, and Yann LeCun on representation learning.

Awards and honors

Doya has received recognition from Japanese and international bodies, including awards and invitations from organizations such as the Japanese Neural Network Society, the Society for Neuroscience, and the IEEE Computational Intelligence Society. He has been a plenary and keynote speaker at meetings like IROS and ICLR satellite workshops, held fellowships from the Japan Society for the Promotion of Science, and served on editorial boards for journals associated with the Society for Neuroscience, IEEE, and publishers such as Nature Publishing Group and Oxford University Press.

Selected publications

- Doya, Kenji. "A Bayesian model of vocal learning in songbirds." In proceedings and journals associated with Society for Neuroscience and conferences like NIPS—work cited alongside authors such as Michael Long and Erik S. Kandel. - Doya, Kenji. "Reinforcement learning in continuous time and space." Papers appearing in venues connected to IEEE and research programs with collaborators from University of Tokyo and RIKEN. - Doya, Kenji. "What are the computations of the cerebellum, the basal ganglia and the cerebral cortex?" Reviews cited in collections published by MIT Press and referenced by researchers including Terrence Sejnowski and Stephan H. Heck. - Doya, Kenji, et al. Edited volumes on computational neuroscience and machine learning, with contributors from DeepMind, Google Brain, and laboratories at Harvard University and Princeton University.

Category:Japanese neuroscientists Category:Computational neuroscientists