A Deep Learning Method for Predicting Low-Cost and High-Accuracy Designing of Anticipatory Layered Nanostructures

Abstract

The production-learning-research-based process is considered as a systematic approach that generates significant scientific and economic values. However, the scientific value often receives greater attention than the economic value, which can render it impractical when the cost of mass production is high. To comprehensively address the requirements of high accuracy and low cost in simulation and fabrication, a deep inverse learning method based on the deep convolutional neural network (CNN) is proposed. This method aims to identify an optimal solution that intelligently utilizes low-cost materials instead of high-cost materials during the training process. By developing an improved training model capable of replacing noble metals with ordinary materials, the proposed approach successfully facilitates the design of lower cost layered nanostructures based on the user requirements. Eventually, it is concluded that the cost of predicted structures by the proposed algorithm is much lower than that of the standard deep CNN. Furthermore, the average relative error of the spectrum is less than 5%. <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$^{}$</tex-math> </inline-formula>