Dynamic Characteristics of LIF Neuron Circuits Using CMOS and Volatile Memristor Devices

Abstract

For several years, the Von Neumann architecture has been the foundation of contemporary computers. The straightforward, less intricate, and easy-to- use nature of this architecture for the processor and memory design, makes this computational framework popular among the other architectures over the past years. The application of this architecture is not sufficient for parallel processing and complex computational tasks. Hence, this has created a huge demand for the architecture that supports extensive parallelism, energy efficiency and scalability. In this regard, researchers are exploring non- Von Neumann architectures like neuromorphic computing systems, which are proved to be potential candidate for the construction of neuron and synapse circuits of the architecture. In this work, CMOS and volatile memristor- based leaky-integrate- fire (LIF) neuron circuits are simulated and analyzed using Cadence Virtuoso. Further, the circuits are compared with respect to attributes such as frequency of spikes, spike duration, pattern of spikes, spiking nature, complexity of the circuit, frequency adaptation, reset circuit requirement, and energy per spike among others. In addition, the work explores and showcase the effect of various input, circuit and memristor device parameters on the spiking nature of the volatile LIF neuron circuit.