Related Experiment Videos
Solving the binding problem of the brain with bi-directional functional connectivity
M Watanabe1, O Nakanishi, K Aihara
1Department of Mathematical Engineering and Information Physics, Graduate School of Engineering, The University of Tokyo, Japan. watanabe@sk.q.t.u-tokyo.ac.jp
Summary
This study introduces a novel neural network model to address the binding problem using dynamic functional connectivity. The model demonstrates how multiple neural assemblies can coexist and be tracked, offering a new perspective on brain information processing.
Area of Science:
- Computational Neuroscience
- Artificial Intelligence
Background:
- The binding problem in neuroscience concerns how the brain integrates disparate sensory information into a unified perception.
- Existing models often struggle to explain dynamic neural interactions and feature integration.
Purpose of the Study:
- To propose a neural network model that solves the binding problem using functional connectivity and bidirectional connections.
- To represent objects as global dynamical cell assemblies organized across multiple neural modules.
Main Methods:
- Developed a neural network with a primary map and two higher modules, each with three bidirectional layers.
- Utilized temporal spike coding and coincidence detector neurons to define dynamic functional connectivity.
- Introduced a three-dimensional Joint-Peri Stimulus Time Histogram (J-PSTH) for tracking cell assemblies.
Main Results:
- The model successfully represents objects as global dynamical cell assemblies.
- Demonstrated that multiple cell assemblies, sharing neurons, can coexist within the network.
- The J-PSTH effectively tracks these dynamic cell assemblies and their changing neuronal constituents.
Conclusions:
- The proposed model offers a viable solution to the binding problem through dynamic functional connectivity.
- The findings suggest a mechanism for flexible and dynamic neural representations in the brain.
- The J-PSTH provides a powerful tool for analyzing complex neural dynamics in computational models.