变分自编码器VAE代码

VAE更倾向于数据生成。当我们训练好了decoder，就可以从标准正态分布生成数据作为解码器的输入，来生成类似但不同于训练数据的新样本，作用类似GAN，用来降维、生成数据。

1.训练代码

import keras
import numpy as np
from sklearn.preprocessing import StandardScaler
from sklearn.preprocessing import MinMaxScaler
import numpy as np
import pandas as pd

# the inital input
x = pd.read_csv("XXX.csv", header=None)
x = np.array(x.values)
#xrd = np.delete(xrd, 0, axis = 1)
#X = ss.fit_transform(X)

# dimension of encode
encoding_dim = 100

# dimension of inital input 
input_X = keras.layers.Input(shape=(901,))
encode = keras.layers.Dense(120, activation='relu')(input_X)
encode = keras.layers.Dense(30, activation='relu')(encode)
encoder_output = keras.layers.Dense(encoding_dim)(encode)


decode = keras.layers.Dense(30, activation='relu')(encoder_output)
decode = keras.layers.Dense(120, activation='relu')(decode)
decoder_output = keras.layers.Dense(901, activation='tanh')(decode)

encoder = keras.Model(inputs=input_X, outputs=encoder_output)
AutoEncoder = keras.Model(inputs=input_X, outputs=decoder_output)

AutoEncoder.compile(optimizer='adam', loss='mse', metrics=['accuracy'])
AutoEncoder.fit(x.T, x.T, epochs=1000, batch_size=1)

# save the weights of encoder
encoder_json = encoder.to_json()
with open('encoder.json','w') as json_file:
    json_file.write(encoder_json)
encoder.save_weights('encoder_weights.hdf5')

# save the weights of vae
AutoEncoder_json = AutoEncoder.to_json()
with open('AutoEncoder.json','w') as json_file:
    json_file.write(AutoEncoder_json)
AutoEncoder.save_weights('vae_weights.hdf5')

2. 测试模型

import os
from keras.models import load_model, model_from_json
import numpy as np

import pandas as pd
x = pd.read_csv("XXX.csv", header=None)
x = np.array(x.values)
## VAE input
model_path = 'AutoEncoder.json'
json_file = open(model_path,'r')
loaded_model_json = json_file.read()
json_file.close()
model = model_from_json(loaded_model_json)
model.load_weights('vae_weights.hdf5')
y = model.predict(x.T)
np.savetxt("YYY.csv", y, delimiter=',')