A Modified Tunable Metal Oxide Memristor Model for Neuromorphic Computing Applications

Abstract:

Memristors are emerging as key building blocks of next-generation computing paradigms, particularly in the field of neuromorphic computing. In this context, a modified Tantalum oxide memristor model, considered here, demonstrates a novel approach for simulating synaptic behavior in neuromorphic systems, with adjustable performance characteristics. The proposed model is implemented using SPICE modeling and simulated using the LTspice simulator. From SPICE modeling, it is observed that the proposed memristor model facilitates the replication of synaptic plasticity, a critical feature for emulating learning and memory processes. Further, the Hodgkin- Huxley Neuristor circuit is designed using the implemented LTspice, where the neuristor shows an action potential amplitude of 0.234 V in response to super-threshold stimuli, indicating effective threshold detection, while sub-threshold inputs lead to voltage attenuation to 0.068 V, enabling precise signal discrimination. Furthermore, the capability of the proposed memristor model is verified by implementing a Neural Network, and it is found that it generates spike potentials in the range of 0.6 V to 0.7 V, analogous to systems with a 0.7 V switching threshold. Also, the data obtained through our calculations show reasonable agreement with experimental IRIS data.