Cortext: a columnar model of bottom-up and top-down processing in the neocortex.

Area of Science:

• Computational Neuroscience
• Systems Neuroscience
• Visual Perception

Background:

• Visual scene recognition involves rapid hypothesis generation (150 ms).
• Mammalian visual cortex employs bidirectional, top-down processing for recognition.

Purpose of the Study:

• To present a spiking neural network model simulating cortical visual processing.
• To demonstrate how the cortex integrates bottom-up and top-down strategies for scene understanding.

Main Methods:

• Developed a spiking neural network model.
• Simulated rapid hypothesis generation and top-down feedback mechanisms.
• Analyzed spike latency to code response reliability.

Main Results:

• Model generates initial hypotheses within milliseconds.
• Relative spike latency indicates response reliability.
• Top-down feedback refines initial hypotheses and deactivates irrelevant representations.

Conclusions:

• Cortical processing integrates rapid gist extraction with top-down refinement.
• Different cortical layers may support distinct processing modes.
• The model provides a framework for understanding visual recognition via predictive coding principles.