Sample-efficient Deep Learning for Surrogate-assisted Evolutionary Optimization of Nanostructures

A Deep Neural Network surrogate-assisted Differential Evolution algorithm for a broadband anti-reflection coating thin-film multilayer optic design. Novel training loss functions, that emphasize a model's ability to predict a structurally similar response, are used. The folder makes available the code, jupyter notebooks, datasets, plotting routines for data display.

本研究提出了一种用于宽带抗反射涂层薄膜多层光学设计的深度神经网络（Deep Neural Network）代理辅助差分进化（Differential Evolution）算法。本算法采用新型训练损失函数，着重强化模型预测结构相似光学响应的能力。本数据集包提供了相关代码、Jupyter笔记本（Jupyter Notebook）、数据集以及用于数据可视化展示的绘图程序。