Prediction model of K2CsSb photocathode reflectivity based on LSTM北大核心CSCD

针对目前K₂CsSb光阴极制备过程中无法预判光阴极生长状态的问题，提出一种基于长短期记忆（LSTM）循环神经网络的K₂CsSb光阴极反射率预测模型。一维原始反射率数据集经过清洗、筛选、序列化等预处理手段后重构为二维数据输入模型。为充分利用反射率数据在时序上高度相关的特性，采用双层LSTM网络提取特征，预测结果通过全连接层输出，以均方误差（MSE）作为模型预测效果的评判标准。实验结果表明，该模型的网络结构合理且在不同数据集下的表现良好，预测准确率可达99.21%。该模型可运用在K₂CsSb光阴极的制作过程中，通过反射率预测值反馈调节工艺参数以趋近目标走势，对提高光阴极性能具有促进作用。

Prediction model of K2CsSb photocathode reflectivity based on LSTM

Aiming at the problem that the growth state of K₂CsSb photocathode cannot be predicted in the current preparation process of K₂CsSb photocathode, a prediction model of K₂CsSb photocathode reflectivity based on long short-term memory (LSTM) recurrent neural network was proposed. The one-dimensional original reflectivity data set was reconstructed into a two-dimensional data input model after cleaning, screening, serialization and other preprocessing methods. In order to make full use of the highly correlated characteristics of reflectivity data in time series, this model used a double-layer LSTM network to extract features, the prediction results were output through the fully connected layer, and the mean square error (MSE) was used as the evaluation standard for the prediction effect of the model. The experimental results show that the network structure of the model is reasonable and performs well in different data sets, and the prediction accuracy rate can reach 99.21%. The proposed model can be used in the fabrication process of K₂CsSb photocathode, and the process parameters can be adjusted by feedback of the reflectivity prediction value to approach the target trend, which can promote the performance of the photocathode.