论文部分内容阅读
采用贝叶斯正则化神经网络(BRNN)对61种金属晶体结合能进行了预测。对网络结构、训练集、预测集以及学习次数进行了优化,并用独立预测样本对贝叶斯正则化神经网络作了检验。预测结果表明,在推广能力方面,贝叶斯正则化神经网络优于熟知的反向传播(BP)神经网络和多元线性回归方法(MLR)。它可望成为元素和化合物构效关系研究的辅助手段。
Sixty-one metal crystal binding energies were predicted using the Bayesian Regularization Neural Network (BRNN). The network structure, training set, prediction set and the number of learning are optimized, and the Bayesian regularization neural network is tested with independent prediction samples. The prediction results show that Bayesian regularization neural network is superior to the well-known backpropagation (BP) neural network and multiple linear regression (MLR) in terms of its promotion ability. It is expected to be an adjunct to the study of the structure-activity relationship of elements and compounds.