Akihiro Funamizu

This paper introduces a hybrid deep reinforcement learning method to model decision-making with multiple strategies, showing it can replicate human behavior in a two-step task.

The paper develops a real-cyber hybrid neural network to predict mouse decision-making movements by incorporating real neuronal activity and generated activity, outperforming traditional ANNs.

The paper identifies localized and global computations of task variables in the cerebral cortex of male mice during decision-making tasks.

This study investigates neural encoding of model-free and inference-based strategies in mice using a tone frequency discrimination task and identifies brain regions involved in reward expectation.

The paper investigates how auditory cortex encodes sound and non-sensory context, finding that neural activity reflects both stimuli and reward expectations, but the decoder remains stable despite variable sensory representations.