Related Experiment Videos
A model of computation in neocortical architecture
E Körner1, M -O. Gewaltig, U Körner
1HONDA R&D Europe (Deutschland) GmbH, Future Technology Research, Carl-Legien-Strasse 30, 63073, Offenbach/Main, Germany
Summary
We propose a new computational model for neocortical function, linking brain architecture to visual recognition. This model uses periodic clocking for discrete processing, differing from standard neural networks.
Area of Science:
- Computational neuroscience
- Cognitive science
- Neurobiology
Background:
- The neocortex's architecture is thought to underlie its computational capabilities.
- Understanding visual recognition involves elucidating how the neocortex processes information.
Purpose of the Study:
- To propose a computational hypothesis linking neocortical architecture to visual recognition.
- To integrate anatomical structures (columns) into a functional framework.
- To introduce a novel model of cortical processing.
Main Methods:
- Development of a large-scale computational hypothesis for visual recognition.
- Incorporation of rapid parallel forward recognition and feedback-controlled refinement.
- Introduction of periodic clocking for discrete cortical processing.
Main Results:
- The model enables parallel categorization and sequential refinement through discrete time steps.
- It presents a functional interpretation of anatomically defined neocortical columns.
- The proposed architecture differs from conventional neural network models.
Conclusions:
- Neocortical architecture is intrinsically linked to its computational principles.
- Periodic clocking provides a mechanism for combining parallel and sequential processing.
- The model offers a new perspective on gamma oscillations and cognitive binding.