Dependency of the connection architectures of oscillatory neural networks on synchronization

Connection architectures of oscillatory neural networks critically affect their computational functions, such as feature-binding/segmentation by synchronous/asychronous oscillation. The synchronous/asynchronous connection types of oscillatory neural networks are quantitatively evaluated by computer simulation. The analytical results are applied to constructing a model of a segmentation task in the visual cortex. It is shown that the orientation features of handwritten numbers are obtained by extracting the synchronous oscillations from the asynchronous background.