Automated Generation of SPICE Models of Memristor-Based Neural Networks from Python Models

Here we propose a new approach to creation of SPICE models of artificial neural networks (ANN) based on memristors. The main feature of this approach is that SPICE models are created automatically based on the information of ANN architecture and structure, extracted from Keras models, as well as the information about parameters of arrays of mem- ristors and hardware neurons specified by an user. Developed software also allows to simulate SPICE models, calculating the nominal values of electronic components automatically. This approach has been developed in order to simplify and speed up the research process and design of ANNs based on memristors.