ECN>> TPU-VAE>> 返回
项目作者： koshian2

项目描述 ：
Variational Auto Encoder in TensorFlow/Keras by TPUs
高级语言： Python
项目地址: git://github.com/koshian2/TPU-VAE.git
创建时间： 2019-01-09T19:44:58Z
项目社区：https://github.com/koshian2/TPU-VAE
开源协议：MIT License
下载


TPU-VAE

A MNIST example of variational auto encoder in TensorFlow/Keras by TPUs.

Overview

Random numbers for sampling and reparameterization trick, and KL divergence of the middle layer are connected by skip-connection, respectively.

This makes it possible to define VAEs in one model, making it easier to train with TPU.