Line Resistance Impact in Memristor-based Multi Layer Perceptron for Pattern Recognition

In this paper we analyse the impact of the line resistance in memristor cross-point arrays (MCA) on the inference accuracy of a Multi-Layer Perceptron (MLP) intended for pattern recognition tasks. Ex-situ training using the MNIST dataset is carried out. The analysis is performed considering different image sizes and a partitioning scheme of the ideal CPA is proposed to minimise the impact of the line resistance. The confusion matrix is used to benchmark the system performance metrics.