vae

代码说明：

  什么是
变分自动编码器? 为了理解VAE，我们首先从最简单的网络说起，然后再一步一步添加额外的部分。一个描述神经网络的常见方法是近似一些我们想建模的函数。然而神经网络也可以被看做是携带信息的数据结构。假如我们有一个带有解卷积层的网络，我们设置输入为值全为1的向量，输出为一张图像。然后，我们可以训练这个网络去减小重构图像和原始图像的平均平方误差。那么训练完后，这个图像的信息就被保留在了网络的参数中。(What is a variational automatic encoder? To understand VAE, we start with the simplest network, and then add additional parts step by step. A common way to describe neural networks is to approximate some of the functions we want to model. However, neural networks can also be regarded as data structures carrying information. If we have a network with deconvolution layer, we set the input to a vector with all values of 1 and output to an image. Then, we can train the network to reduce the average square error between the reconstructed image and the original image. After training, the information of the image is retained in the parameters of the network.vae)

文件列表：

vae.py, 4549 , 2019-01-27