Stability for delayed reaction-diffusion neural networks

We consider a Hopfield neural network model with diffusive terms, non-decreasing and discontinuous neural activation functions, time-dependent delays and time-periodic coefficients. We provide conditions on interconnection matrices and delays which guarantee that for each periodic input the model has a unique periodic solution that is globally exponentially stable. Even in the case without diffusion, such conditions improve recent results on classical delayed Hopfield neural networks with discontinuous activation functions. Numerical examples illustrate the results.