Learning temporal sequences by excitatory synaptic changes only

We considers fully connected neural networks where neurons are divided into two groups: inhibitory neurons whose outgoing synapses are all inhibitory, and excitatory neurons, the outgoing synapses of which are all excitatory. We show that such networks are capable of learning temporal sequences. Our model is similar to one previously described by us, but here learning takes place only in the excitatory-to-excitatory synapses. Mean-field equations of the model are presented. Numerical solution and Monte Carlo simulation demonstrate model performance.