Multiscale representations of community structures in attractor neural networks

Autor: Tomoki Fukai, Tatsuya Haga
Rok vydání: 2021
Předmět:
DYNAMICS
Theoretical computer science
Computer science
Social Sciences
Hippocampus
Infographics
Cognition
Learning and Memory
Attractor
Medicine and Health Sciences
Psychology
Biology (General)
Theta Rhythm
NEURONS
Associative property
Attractor network
Data Management
Random graph
Artificial neural network
Ecology
CIRCUIT
Brain
Content-addressable memory
Computational Theory and Mathematics
Community Ecology
Modeling and Simulation
Physical Sciences
Anatomy
Graphs
Network Analysis
STORAGE
Research Article
Computer and Information Sciences
QH301-705.5
Models
Neurological
Cellular and Molecular Neuroscience
Memory
Genetics
Humans
Learning
Molecular Biology
Community Structure
Ecology
Evolution
Behavior and Systematics
FINE INFORMATION
Data Visualization
Ecology and Environmental Sciences
Cognitive Psychology
Biology and Life Sciences
Graph theory
Eigenvalues
MODEL
Algebra
Linear Algebra
Cognitive Science
Neural Networks
Computer
Laplacian matrix
Eigenvectors
Mathematics
Neuroscience
Zdroj: PLoS Computational Biology
PLoS Computational Biology, Vol 17, Iss 8, p e1009296 (2021)
ISSN: 1553-7358
Popis: Our cognition relies on the ability of the brain to segment hierarchically structured events on multiple scales. Recent evidence suggests that the brain performs this event segmentation based on the structure of state-transition graphs behind sequential experiences. However, the underlying circuit mechanisms are poorly understood. In this paper we propose an extended attractor network model for graph-based hierarchical computation which we call the Laplacian associative memory. This model generates multiscale representations for communities (clusters) of associative links between memory items, and the scale is regulated by the heterogenous modulation of inhibitory circuits. We analytically and numerically show that these representations correspond to graph Laplacian eigenvectors, a popular method for graph segmentation and dimensionality reduction. Finally, we demonstrate that our model exhibits chunked sequential activity patterns resembling hippocampal theta sequences. Our model connects graph theory and attractor dynamics to provide a biologically plausible mechanism for abstraction in the brain.
Author summary Our experiences are often hierarchically organized, so is our knowledge. Identifying meaningful segments in hierarchically structured information is crucial for many cognitive functions including visual, auditory, motor, memory, language processing, and reasoning. Herein, we show that the attractor dynamics of recurrent neural circuits offer a biologically plausible way for hierarchical segmentation. We found that an extended model of associative memory autonomously performs segmentation by finding groups of tightly linked memories. We proved that the neural dynamics of our model mathematically coincide with optimal graph segmentation in graph theory and are consistent with the experimentally observed nature of human behaviors and neural activities. Our model established a previously unexpected relationship between attractor neural networks and the graph-theoretic processing of knowledge structures. Our model also provides experimentally testable predictions, particularly regarding the role of inhibitory circuits in controlling representational granularity.
Databáze: OpenAIRE
Externí odkaz:
Nepřihlášeným uživatelům se plný text nezobrazuje