Synthesis and Analysis of a Memristor-Based Artificial Neuron

The neural networks and their basic components - the artificial neurons are very important modules in electronics. Owing to their extensive usage, it is of high interest their new schematic solutions to be investigated. The objective of this research is to propose a complete analysis of a suggested memristor-based linear neuron with memristorbased synapses. The analyzed circuit is based on a classical linear neuron for noise cancellation and a resistor-memristor synaptic circuit. The applied memristor synapse is capable to realize positive, zero and negative synaptic weights. For the simulations, a previously proposed by the author in another paper modified nonlinear memristor model is used. Several basic memristor models are also applied. A comparison between the results is realized. Advantages of the proposed circuit are the comparatively wide range of changing the synaptic weights, their simple tuning process by voltage pulses and the use of only one memristor element.