Towards improving generalization of a neural network by interpretation for topological phases of matter

Machine learning (ML) promises a revolution in science, similarly as it has already revolutionized our everyday lives. In quantum physics, this tool is especially promising in the detection of phases of matter. However, ML models are also known for their black-box construction, which hinders understanding of what they learn from the data and makes their application to novel data risky. Moreover, the central challenge of ML is to ensure its good generalization abilities, i.e., good performance on data outside the training set. Here, we show how the informed use of an interpretability method called class activation mapping (CAM) and its extensions increases the reliability of a neural network (NN) trained to classify quantum phases. In particular, we show that we can ensure better generalization in the complex classification problem by choosing such a model that, in the simplified version of the problem, learns a known characteristics of the phase. We show this on an example of the topological Su–Schrieffer–Heeger (SSH) model with and without the disorder. This work is an example of how the routine use of interpretability methods can improve the performance of ML in scientific problems.