The Cooperative Network Architecture: Learning Structured Networks as Representation of Sensory Patterns
This addresses problems in vision systems for AI/robotics by offering a new approach to invariant object recognition.
The paper tackles the challenge of representing sensory signals by introducing the Cooperative Network Architecture (CNA), which uses dynamically assembled, recurrently connected networks of neurons to encode patterns, resulting in robustness to noise and generalization to out-of-distribution data.
We introduce the Cooperative Network Architecture (CNA), a model that represents sensory signals using structured, recurrently connected networks of neurons, termed "nets." Nets are dynamically assembled from overlapping net fragments, which are learned based on statistical regularities in sensory input. This architecture offers robustness to noise, deformation, and generalization to out-of-distribution data, addressing challenges in current vision systems from a novel perspective. We demonstrate that net fragments can be learned without supervision and flexibly recombined to encode novel patterns, enabling figure completion and resilience to noise. Our findings establish CNA as a promising paradigm for developing neural representations that integrate local feature processing with global structure formation, providing a foundation for future research on invariant object recognition.