实现对于数据特征的识别和提取，进而实现重构，重构值与输入数据相差越小越好，对于该深度学习网络的训练可以分为预训练和调优过程，对此需要把数据按比例分为训练集和测试集，进行系数的调整，从而实现数据重构。-For realization identify and extract data features, thus achieving the reconstruction, reconstruction of the input data value difference as small as possible, for the training of the deep learning network can be divided into pre-training and tuning process, which requires the data to scale divided into a training set and test set, adjustment factor, enabling data reconstruction.