Autors: Mladenov, V. M., Tsenov, G. T., Kirilov, S. M.
Title: LTSPICE Memristor Neuron with MOS Transistor-Based Logarithmic-Sigmoidal Activation Function
Keywords: LTSPICE; MATLAB; memristor-based neuron; memristor-based syn

Abstract: Memristors are novel electronic nano-sized elements with very good switching and memory properties. Owing to their low power consumption and a good compatibility to present CMOS high-density integrated circuits, they could be applied in neural networks, memory arrays, and many other electronic devices. In this paper, simple and modified LTSPICE models of a memristor, an activation function and artificial neuron with memristor synapses are suggested. In neuron's realization, each synaptic weight is implemented by only one memristor, proposing a massively reduced schematic complexity. The scaling and summing circuits implementation is founded on memristors and op-amps. The proposed logarithmic-sigmoidal activation function is implemented by a simple circuit with MOS transistors and memristors. The operation of the suggested memristor neuron is analyzed in MATLAB and LTSPICE and the obtained results are compared and verified. The proposed memristor-based artificial neuron is an important

References

    Issue

    CNNA, pp. 1-4, 2023, Greece, IEEE, doi: 10.1109/CNNA60945.2023.10652671

    Вид: публикация в международен форум, публикация в реферирано издание, индексирана в Scopus