说明：  autoencoder是一种无监督的学习算法，主要用于数据的降维或者特征的抽取，在深度学习中，autoencoder可用于在训练阶段开始前，确定权重矩阵的初始值。
(Autoencoder is an unsupervised learning algorithm, which is mainly used for dimensionalization reduction or feature extraction of data. In deep learning, autoencoder can be used to determine the initial value of weight matrix before the training stage.)