This code is a tensorflow implementation of the algorithm on the paper: Autoencoder with Invertible Functions for Dimension Reduction and Image Reconstruction
Author of this code: Peizhi Yan
Affiliation: Lakehead University
Personal Website: https://PeizhiYan.github.io
Date: March 24th, 2019
Tensorflow version: r1.13
For example, assume you want to reduce the dimension of MNIST dataset from 28*28 to 100, you can use the following lines of code:
"""create the ELA model"""
model = ELA(input_units=28*28, hidden_units=100, c=2**(8), activation='sin')
"""train the model"""
model.train(x_train, 100) # x_train is the training dataset with shape [50000,28*28]; 100 is the numer of total training epochs
"""evaluate the model"""
print("MSE(train) :", model.evaluate(x_train))
print("MSE(train) :", model.evaluate(x_test))
"""encode data"""
x_train_encoding = model.encoding(x_train)
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent