摘要
本文用ＤＴＡ法测量了Ｇａ－Ｉｎ－Ａｓ－Ｓｂ系合金的液相线温度，并用滑舟法在Ｇａｓｈ衬底上生长液相外延膜，并用电子探针法分析膜的成分．在此基础上利用实验数据和文献数据训练人工神经网络模型，神经网络模型用交叉检验法测试是否过拟合．经测试合格的神经网络可以预报液相线温度和外延层组成．
Abstract Measurement and computer prediction of the Ga-In-As-Sb solid-liquid equilibrium are reported. The liquidus data are obtained by DTA, and the solidus data are determined by liquid phase epitaxial method. The artificial neural networks,trained by use of experimental data,can be used to predict liquidus and solidus data. And cross-validation method is also employed to get rid of overfitting.
