Stability and Intermittency in Large-Scale Coupled Oscillator Models for Perceptual Segmentation

Area of Science:

• Computational neuroscience
• Artificial intelligence
• Dynamical systems

Background:

• Perceptual segmentation is crucial for visual processing.
• Understanding the dynamics of pattern formation in neural systems is key.
• Coupled map lattices offer a framework for modeling complex neural dynamics.

Purpose of the Study:

• To propose a coupled map lattice as a model for perceptual segmentation.
• To investigate the role of chaotic dynamics in pattern formation and stability.
• To explore how adaptive connections and oscillations influence segmentation.

Main Methods:

• Utilizing a coupled map lattice with locally coupled nonlinear maps.
• Analyzing high-dimensional, deterministic chaos to generate synchronized activity patterns.
• Employing analytical tools and numerical simulations to study stability characteristics.
• Introducing adaptive connections and stimulus-controlled oscillations.

Main Results:

• Chaotic dynamics contribute to pattern creation and destabilize synchronized states.
• Network coupling strength and chaotic divergence rate determine stability.
• Stable or meta-stable patterns emerge with adaptive connections and oscillations, reflecting visual field structure.
• The model replicates experimental switching-time distributions for ambiguous patterns.

Conclusions:

• Coupled map lattices can model perceptual segmentation through synchronized chaotic dynamics.
• The model provides insights into the stability and adaptability of neural pattern formation.
• The system's behavior, including switching between segmentations, aligns with psychological observations.