Memorizing morph patterns in small-world neuronal network

In this paper, we study the memory representation of morph patterns in an attractor neural network model. Since recent studies indicate that biological neural networks exhibit the so-called small-world effect, we study here how the small-world connection topology affects the dynamics of memory representation of morph patterns. We find that the small-world connection has significant effects on the memory representation dynamics in the network. Based on this finding, we postulate that global (or long-range) synaptic connections are mainly responsible for learning patterns that are significantly different from those already stored. Further numerical simulations show that the model based on this hypothesis has several advantages, for example fast learning and good performance.