An important, but unresolved question with regards to kirigami structures at all length scales is how to locate the cuts to achieve a specific performance metric. This problem is challenging to solve due to the large numbers of possible cut configurations that must be explored. 剪纸结构的大数据来源，需要确定Cuts（切口）

Thus, exhaustively searching for good solutions in this design space would be impractical as the number of possible configurations grows exponentially with the system size. Alternatively, various optimization algorithms, i.e., genetic and greedy algorithms, and topology optimization approaches, have been used to find optimal designs of materials based on finite element methods [17–20]. However, these approaches have difficulties as the number of degrees of freedom in the problem increases, and also if the property of interest lies within the regime of nonlinear material behavior. 写的有逻辑，可以借鉴句式

We use fully connected neural networks (NNs) and also convolutional neural networks (CNNs) to approximate the yield strain and stress. 看到很多次NNs与approximate这个词的搭配了，predict很少见到

Our findings can be used as a general method to design a material without any prior knowledge of the fundamental physics, which is particularly important for designing materials when only experimental data are available and an accurate physical model is unknown. 如果数据可以实验得到，那么可以理解为不需要任何物理先验知识