Analytical model for memristive systems for neuromorphic computation. (17th June 2021)

• Record Type: Journal Article
• Authors: Kumar, Sanjay; Agrawal, Rajan; Das, Mangal; Jyoti, Kumari; Kumar, Pawan; Mukherjee, Shaibal
• Is Part Of: Journal of physics. Volume 54:Number 35(2021)
• Publication Date: 2021-06-17
• Subjects: artificial neuron -- modeling -- neuromorphic computation -- memristive systems; Physics -- Periodicals; 530
• DOI: 10.1088/1361-6463/ac07dd
• Languages: English
• ISSNs: 0022-3727

Abstract

Abstract: From the last decade, the development of a generic model for memristive systems which simulates the biologically inspired nervous system of living beings, is one of the most attracting aspects. More specifically, the develop generic model has capability to resolve the problems in the field of artificial neural network. Here, a generic, non-linear analytical memristive model, which is based on interfacial switching mechanism, has been discussed. The proposed model has the capability to simulate the high-density neural network of biological synapses that regulates the communication efficacy among neurons and can implement the learning capability of the neurons. Further, the proposed model is the parallel connection of the rectifier and memristor which shows better non-linear profile along with non-ideal effects and rectifying nature in its pinched hysteresis loop in the resistive switching characteristics. Moreover, proposed model shows the significant low value of maximum error deviation ∼4.44% for Y2 O3 -based and ∼4.5% for WO3 -based memristive systems, respectively in its neuromorphic characteristics with respect to the corresponding experimental results. Therefore, the proposed analytical memristive model can be utilized to develop the memristive system for real-world applications based on neuromorphic behaviors of any transition metal oxide-based memristive systems.