The skin effect, which is the extreme sensitivity of the spectrum and eigenstates to the boundary condition, is a remarkable phenomenon of non-Hermitian systems and is currently being actively studied. In particular, the mirror skin effect, which is the protection of the skin effect by mirror symmetry, has recently been discovered. In this paper, we clarify the robustness of the mirror skin effect against disorder that disrupts the mirror symmetry. Specifically, we employed a neural network to elucidate the robustness of the skin effect. The neural network is useful because it systematically predicts the presence/absence of skin modes in the form of a large number of localized states around the edge. The trained neural network detects skin effects with high accuracy and clarifies the phase diagram of the model. We also calculate the probabilities the neural network obtained for each of the states. The above results were additionally confirmed by calculating the inverse participation ratio.

中文翻译：

---

机器学习中存在障碍的镜面皮肤效应

趋肤效应是光谱和本征态对边界条件的极端敏感性，是非Hermitian系统的显着现象，目前正在积极研究中。特别地，最近发现了镜面皮肤效应，该镜面皮肤效应是通过镜面对称来保护皮肤效果的。在本文中，我们阐明了镜面皮肤对抵抗破坏镜面对称性的疾病的鲁棒性。具体来说，我们采用了神经网络来阐明皮肤效应的鲁棒性。神经网络之所以有用，是因为它以边缘附近的大量局部状态的形式系统地预测了皮肤模式的存在/不存在。经过训练的神经网络可以高精度检测皮肤效应，并阐明模型的相图。我们还计算了神经网络为每个状态获得的概率。通过计算逆参与率，可以进一步确认上述结果。