Object separation in dynamic neural networks

It has been proposed that correlated neural activity is a functional principle for feature linking and object separation in the visual system. The results of the authors´ neural network simulations support this hypothesis. Of particular interest is the fact that a simple neural network without feedback from an associative memory is able to perform a scene segmentation task on the basis of object domain data only. Simulations with moving objects show that object definition via synchronous ensemble activity is maintained over a considerable velocity range