自动编码线性解码器主要是考虑到稀疏自动编码器最后一层输出如果用sigmoid函数，由于稀疏自动编码器学习是的输出等于输入，simoid函数的值域在[0,1]之间，这就要求输入也必须在[0,1]之间，这是对输入特征的隐藏限制，为了解除这一限制，我们可以使最后一层用线性函数及a z-Automatic linear encoder decoder mainly on account of sparse autocoder last layer output if sigmoid function, since learning sparse automatic encoder output is equal to input range simoid function between [0,1], which must also be asked to enter [0,1], which is the input feature hidden restrictions, in order to lift this restriction, we can make the last layer of a linear function and a　 z