神经网络学习对于逼近实数值、离散值或向量值的目标函数提供了一种健壮性很强的方法 -Neural network learning approach for real values, discrete values ​ ​ or provide a strong method to measure the robustness of the objective function