Attractor-Based Models for Sequences and Pattern Generation in Neural Circuits.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Theoretical Neuroscience

Background:

• Neural circuits are crucial for functions like locomotion and breathing, often involving rhythmic activity.
• Traditional models use coupled oscillators for rhythm generation and attractor networks for pattern classification.
• Existing models struggle to unify these distinct neural functions.

Purpose of the Study:

• To present a unified theoretical framework for attractor-based neural networks.
• To demonstrate how attractor networks can generate diverse rhythmic patterns, including those of central pattern generator circuits (CPGs).
• To propose a mechanism for pattern transition and sequence generation.

Main Methods:

• Developed a theoretical framework using threshold-linear networks.
• Constructed a network with two distinct modules: a counter network (fixed points) and a locomotion network (limit cycles).
• Introduced a novel layered network architecture producing fusion attractors to link network modules.

Main Results:

• Demonstrated that attractor-based networks can generate diverse rhythmic patterns.
• Showcased a network capable of sequencing through five different quadruped gaits.
• Successfully linked a counter network to a locomotion network to achieve sequential gait transitions.

Conclusions:

• Attractor networks provide a unified framework for both pattern classification and rhythmic pattern generation.
• The proposed layered architecture and fusion attractors enable complex sequential behaviors, like quadruped locomotion.
• This work advances computational models of neural function, bridging previously separate modeling approaches.