项目作者: snehalvartak

项目描述 :
Fully connected network on MNIST data using Tensorflow
高级语言: Jupyter Notebook
项目地址: git://github.com/snehalvartak/MNIST.git
创建时间: 2018-03-24T18:42:06Z
项目社区:https://github.com/snehalvartak/MNIST

开源协议:

下载


MNIST Classification using Deep Neural Network

Here I have trained a fully connected net for MNIST Classification using Tensorflow.
This fully connected network has 5 hidden layers with 1024 hidden units each.
To initialize the weights I have used Xavier initialization technique taught in class.
This initialization coupled with batch normalization and Adam Optimizer gives an accuracy of 98.3% on test data.