2017 Fiscal Year Annual Research Report
Using recurrent neural networks to study neural computations in cortical networks
Publicly Offered Research
| Project Area | Correspondence and Fusion of Artificial Intelligence and Brain Science |
|---|
| Project/Area Number |
17H06037
|
|---|
| Research Institution | Institute of Physical and Chemical Research |
|---|
Principal Investigator |
ベヌッチ アンドレア 国立研究開発法人理化学研究所, 脳神経科学研究センター, チームリーダー (50722352)
|
|---|
| Project Period (FY) |
2017-04-01 – 2019-03-31
|
| Keywords | Decision-making / sensory processing / visual cortex / neural computation / optogenetics / cognition |
|---|
| Outline of Annual Research Achievements |
The research carried over the fiscal year FY2017 has led to the refinement and optimization of computational tools for the analysis of large-scale neural recordings and to advance our understanding on fundamental coding questions on sensory processing and sensory-based decision making. In regards to the tool development, we have progressed on experimental frameworks for widefield data from the mouse occipital cortex during goal-directed behaviors. Recordings consisted on GCaMP signals across 10-12 visual cortical areas as the mouse performed in a two-alternative forced choice orientation discrimination task. This was done with cell-type specificity for excitatory and inhibitory neurons in superficial layers (L2/3) of the mouse visual cortex, using an in-vivo awake preparation with two-photon and widefield calcium imaging. Combined with a digital micromirror display (DMD) we routinely performed optogenetic perturbations of the network dynamics at near single-cell resolution. We have analyzed the data with linear and non-linear dimensionality reduction methods and managed to identify central task variables in complex correlational interdependencies between the activities of occipital-parietal visual areas.
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
we did not encounter any problem
|
| Strategy for Future Research Activity |
The research plan for next fiscal year focuses on theoretical modeling of the experimental data using supervised and unsupervised methods to reveal the latent dynamics in widefield and two-photon data. In particular, the analysis will attempt to reveal complex task signatures besides the immediate sensory-related ones and to establish a formalism for a psychometric to neurometric mapping. For example, the behavioral task described in the proposal revolves around the animal’s cognitive ability to form a perceptual concept, defined as the ability to perform in a categorization task with perceptual invariances (size, rotation, contrast, spatial frequency, etc.). The analysis will strive to highlight neural signatures linked to such invariances.
|
