How to see hidden patterns in metamaterials with interpretable machine learning

Machine learning models can assist with metamaterials design by approximating computationally expensive simulators or solving inverse problems. However, past work has usually relied on black box deep neural networks, whose reasoning processes are opaque and require enormous datasets that to obtain. In this work, we develop two novel machine approaches discovery have neither of these disadvantages. These approaches, called shape-frequency features unit-cell templates , discover 2D user-specified frequency band gaps. Our provide logical rule-based conditions metamaterial unit-cells allow for interpretable processes, generalize well across spaces different resolutions. The also flexibility where users almost freely the fine resolution a without affecting user’s desired gap.