Heterogeneous Coexisting Attractors and Large-Scale Amplitude Control in a Simple Memristive Neural Network

This paper proposes a simple ring memristive neural network (MNN) with self-connection, bidirectional connection and a single memristive synapse. Compared with some existing MNNs, the most distinctive feature of the proposed MNN is that it can generate heterogeneous coexisting attractors and large-scale amplitude control. Various kinds of heterogeneous coexisting attractors are numerically found in the MNN, including chaos with a stable point, chaos with a limit cycle, a limit cycle with a stable point. By increasing the parameter values, the chaotic variables of the MNN can be accordingly increased and their corresponding areas are extremely wide, yielding parameter-dependent large-scale amplitude control. A circuit implementation platform is established and the obtained results demonstrate its validity and reliability.

Automated summary

This paper proposes a simple ring memristive neural network (MNN) with unique features of generating heterogeneous coexisting attractors and enabling large-scale amplitude control. Various types of attractors are numerically found in the MNN, including chaos with a stable point, chaos with a limit cycle, and a limit cycle with a stable point. The MNN's chaotic variables can be increased by adjusting parameter values, allowing for parameter-dependent large-scale amplitude control. A circuit implementation platform is established and experimental results demonstrate the validity and reliability of the proposed MNN.