A great success in detecting the phase transition in a two-dimensional Ising model using a neural network prompts us to apply the idea of nonequilibrium relaxation to the detection. In fact, convolutional neural networks can afford to predict the transition point at an early learning stage. We also mention the limitations of the machine learning approach.

中文翻译：

---

非平衡松弛方案在机器学习中检测相变的应用

在使用神经网络检测二维伊辛模型中的相变方面取得了巨大的成功，这促使我们将非平衡弛豫的思想应用到检测中。实际上，卷积神经网络可以在早期学习阶段预测过渡点。我们还提到了机器学习方法的局限性。