Single neuron learns temporal pattern via memristive STDP with pulse-width modulation

In this work, we investigate the possibility of learning in spiking neural network with connections based on the model of Au/Ta/ZrO2(Y)/Ta2O5/TiN/Ti memristive devices. These memristive connections demonstrate STDP obtained by pulse-width modulation. We demonstrate that the output neuron successfully learns a temporal pattern by self-organization of memristive connections with proposed STDP mechanism. The output neuron is capable to learn a pattern interspersed with noise due to memristive STDP with pulse-width modulation. We have noted that the efficiency of neuron learning depends on the frequency of pattern presentation in this model, which is due to the fact that the model is based on memristive devices with spike-rate-dependent plasticity. The obtained results suggest that similar self-learning process could be implemented using physical memristive devices and allow to develop neuromorphic technologies that implement the principles of energy-efficient information processing.