自旋霍尔电导与反常霍尔电导材料电导性质预测研究

Prediction of conductivity properties of spin Hall conductive and anomalous Hall conductive materials

将机器学习技术与材料科学领域的相关研究进行结合,以达成材料研究的最佳结果,这是今后研究的大势所趋。本研究针对自旋霍尔电导与反常霍尔电导材料电导特性,提出了一种新的基于PSO-BPNN的元器件导电性能预测模型。实验结果表明,该模型的准确率比BPNN模型有了较大的提高,MSE、RMSE、MAE和R~2四项指标分别提高了51.8%、69.5%、71.2%和50.3%。同时,模型训练时间也比BPNN模型提高了28.19%,四组数据集的拟合性也得到了明显的变化。总之,自旋霍尔电导和反常霍尔电导的电导性质研究是一个重要的热点,因此,研究者还需要探索更多更有效的方法来预测材料的电导性质,以促进自旋霍尔电导和反常霍尔电导材料在高速低功耗电子器件中的应用。

Combining machine learning techniques with related research in the field of materials science to reach the best results in materials research is a major trend in future research.In this study,a new PSO-BPNN-based component conductivity prediction model is proposed for the conductivity properties of spin Hall conductive and anomalous Hall conductive materials.The experimental results show that the accuracy of the model has improved significantly over the BPNN model,with the four metrics of MSE,RMSE,MAE and R2 improving by 51.8%,69.5%,71.2% and 50.3% respectively.The model training time was also improved by 28.19% over the BPNN model,and the fit of the four data sets was significantly changed.In conclusion,the study of the conductivity properties of spin Hall conductivity and anomalous Hall conductivity is an important hotspot,and therefore researchers still need to explore more and more effective methods to predict the conductivity properties of the materials in order to promote the application of spin Hall conductivity and anomalous Hall conductivity materials in high-speed low-power electronic devices.