“Dynamic circuit shift depending on behavioral strategies”

In uncertain environments, we often cannot get enough sensory inputs to know the current context. Therefore we must infer it by simulating how it changes by our own actions or external factors. One illustrative example of this mental simulation is a Mexican game piñata, in which a person tries to hit a piñata far away with eyes closed. Person estimates the distance to piñata from uncertain sensory inputs (i.e., words from others) and own actions.

Psychophysics studies show that such context estimations are achieved by Bayesian inference in humans. Bayesian inference estimates the posterior belief of context by combining its prior and sensory observations (likelihood). The posterior belief is then employed as the prior in next-step estimation by utilizing mental simulation. Bayesian inference is hypothetically implemented in the cerebral cortex which integrates sensory bottom-up signals and predictive top-down signals. However, this hypothesis is mainly based on anatomical evidences, and is almost not tested with the neuronal activities.

Here, we investigate the cortical implementation of Bayesian inference by in-vivo two-photon microscopy and optogenetic manipulation of neurons in mice. Especially, we focus on how the cortical network shifts with sensory- or mental-simulation-based context estimations. We use a genetically encoded calcium indicator, GCaMP6f, in posterior parietal cortex (PPC) of mice to detect neuronal activities. We use channelrhodopsin-2 (ChR2) or archaerhodopsin-T (ArchT) for activating or inactivating neurons. Our study thus combines experimental and theoretical approaches. We hope that our study contributes to not only understand the mechanism of cortical computation but also develop cortex-inspired machine learning algorithms such as deep convolutional neural networks (DCNN).

 
Recent Publications
1. (in Japanese) 船水章大, 銅谷賢治 (2015) 予測―大脳新皮質のベイジアンフィルタ仮説. 生体の科学 66: 33-37.
2. Funamizu A, Ito M, Doya K, Kanzaki R, Takahashi H (2015) Condition interference in rats performing a choice task with switched variable- and fixed-reward conditions. Front Neurosci 9: 00027.
3. Funamizu A, Kanzaki R, Takahashi H (2013) Pre-attentive context-specific representation of fear memory in the auditory cortex of rat. Plos One 8: e63655.
4. Funamizu A, Ito M, Doya K, Kanzaki R, Takahashi H (2012) Uncertainty in action-value estimation affects both action choice and learning rate of the choice behaviors of rats. Eur J Neurosci 35:1180-1189.

Posted:2016/03/10