Neural Networks: A Comprehensive Foundation
Neural Networks: A Comprehensive Foundation
Distortion Invariant Object Recognition in the Dynamic Link Architecture
IEEE Transactions on Computers
Complex-Valued Neural Networks (Studies in Computational Intelligence)
Complex-Valued Neural Networks (Studies in Computational Intelligence)
Temporal coding: Assembly formation through constructive interference
Neural Computation
Temporal coding: competition for coherence and new perspectives on assembly formation
IJCNN'09 Proceedings of the 2009 international joint conference on Neural Networks
| Hi-index | 0.01 |
A pattern recognition mechanism (''windowing mechanism'') is proposed that is based on Hebbian memory and oscillatory inhibition. It is related to properties of cortical gamma oscillations as expressed with the communication-through-coherence hypothesis. The rhythm dominating among excitatory units is imprinted on the inhibitory units through columnar couplings. Excitatory units that participate in the dominating rhythm may escape the feedback from the inhibitory pool through being active in the time windows given by the minima of the inhibitory effect. The activity of other units is suppressed. Constituting an excitatory rhythm that is compatible with the inhibitory one, the participating units comprise the winning patterns. This windowing mechanism is specified through giving an oscillatory network model and demonstrating it with image processing examples. Thereby, we also provide a gradient system formulation for inhibitory generation of synchrony.