How does the brain perceive and interpret sensory inputs to perform behavior? How does the prior knowledge affect sensory perception? Our lab identifies the neuronal processes of decision making and compares the functions with machine learning algorithms.

About Research

Understand the brain mechanism of decision making through machine learning

Our laboratory studies the circuitry of decision making in the dorsal cortices. Our aim is to understand how the brain generates complex behaviors by combining sensory inputs and prior knowledge. We are particularly interested in how the neuronal processes of decision making are functionally different from machine learning algorithms. Although recent artificial neural networks (or artificial intelligence: AI) achieve magnificent performance in visual processing, Shogi, Go, and StarCraft, there are still some tasks which are easy to solve for animals but difficult for AI. We use mice as a model system and combine behavioral tasks, calcium imaging, optogenetics, electrophysiology, and computations to address these questions.

My bachelor’s degree is in engineering, especially in robotics, from the University of Tokyo. At that time, before the era of deep learning, I got interested in neuroscience to understand how the brain accomplishes the sophisticated sensory processes and action selection. After I got a PhD in information science and technology, I continued my career in neuroscience with Dr. Kenji Doya at Okinawa Institute of Science and Technology (OIST) and learned interdisciplinary approaches of computational theory and experiment. At OIST, I investigated the neural substrate of dynamic Bayesian inference in the cerebral cortex. I then did my second postdoc with Dr. Anthony Zador at Cold Spring Harbor Laboratory (CSHL) and studied the neural substrate of perceptual decision making in the mouse auditory cortex.

Our laboratory combines machine learning and animal experiments to understand the neural substrate of decision making. We succeeded to model mouse behavior with reinforcement learning and Bayesian inference, and decode mice position from population neuronal activity. We image calcium signals from dorsal cortical neurons and manipulate them with optogenetics.


  1. Funamizu A, Integration of sensory evidence and reward expectation in mouse perceptual decision-making task with various sensory uncertainties. iScience volume 24, 102826 (2021)
  2. Funamizu A, Kuhn B, Doya K. Neural substrate of dynamic Bayesian inference in the cerebral cortex. Nature Neuroscience volume 19, pages1682–1689(2016)
  3. Funamizu A, Ito M, Doya K, Kanzaki R, Takahashi H. Condition interference in rats performing a choice task with switched variable- and fixed-reward conditions. Frontiers in Neuroscience (2015)
  4. Funamizu A, Kanzaki R, Takahashi H. Pre-Attentive, context-specific representation of fear memory in the auditory cortex of rat. Plos One , 6;8(5):e63655(2013)
  5. Funamizu A, Ito M, Doya K, Kanzaki R, Takahashi H. Uncertainty in action-value estimation affects both action choice and learning rate of the choice behaviors of rats. European Journal of Neuroscience, 2012 Apr; 35(7): 1180–1189.
  6. Funamizu A, Kanzaki R, Takahashi H. Distributed representation of tone frequency in highly decodable spatio-temporal activity in the auditory cortex. Neural Networks, 24(2011)321-332.
Akihiro Funamizu
Graduate School of Arts and Sciences
Kotaro Ishizu
Research Associate