StackedAeModel.py

import keras
from keras.models import Model,Sequential
from keras.layers import Input,Dense, Dropout, Activation, Flatten, Reshape, Conv2D, MaxPooling2D, AveragePooling2D
from keras import regularizers
from keras.losses import mean_squared_error
from keras import losses
import matplotlib.patches as patches
import numpy as np
import dicom
import cv2
import matplotlib.pyplot as plt

def model1(X_train, get_history=False, verbose=0, param_reg=0.001):
    autoencoder_0 = Sequential()
    encoder_0 = Dense(input_dim=4096, units=100, kernel_regularizer=regularizers.l2(param_reg))
    decoder_0 = Dense(input_dim=100, units=4096, kernel_regularizer=regularizers.l2(param_reg))
    autoencoder_0.add(encoder_0)
    autoencoder_0.add(decoder_0)
    autoencoder_0.compile(loss= 'mse',optimizer='adam', metrics=['accuracy'])
    h = autoencoder_0.fit(X_train, X_train, epochs=200, verbose=verbose)

    temp_0 = Sequential()
    temp_0.add(encoder_0)
    temp_0.compile(loss= 'mse', optimizer='adam', metrics=['accuracy'])
    encoded_X = temp_0.predict(X_train, verbose=0)
    if get_history:
        return h.history['loss'], encoded_X, encoder_0
    else:
        return encoded_X, encoder_0

def model2(X_train,encoded_X, encoder_0, get_history=False, verbose=0, param_reg=0.001):
    autoencoder_1 = Sequential()
    encoder_1 = Dense(input_dim=100, units=100, kernel_regularizer=regularizers.l2(param_reg))
    decoder_1 = Dense(input_dim=100, units=100, kernel_regularizer=regularizers.l2(param_reg))
    autoencoder_1.add(encoder_1)
    autoencoder_1.add(decoder_1)
    autoencoder_1.compile(loss= 'mse', optimizer='adam', metrics=['accuracy'])
    h = autoencoder_1.fit(encoded_X, encoded_X, epochs=200, verbose=verbose)

    temp_0 = Sequential()
    temp_0.add(encoder_0)
    temp_0.compile(loss= 'mse', optimizer='adam', metrics=['accuracy'])
    encoded_X = temp_0.predict(X_train, verbose=0)
    if get_history:
        return h.history['loss'], encoder_1
    else:
        return encoder_1

def model3(X_train, Y_train, encoder_0, encoder_1, init='zero',
           get_history=False, verbose=0, param_reg=0.001):
    model = Sequential()
    model.add(encoder_0)
    model.add(encoder_1)
    model.add(Dense(input_dim=100, units=4096, kernel_initializer=init, kernel_regularizer=regularizers.l2(param_reg)))
    model.compile(optimizer = 'adam', loss = "MSE", metrics=['accuracy'])
    h = model.fit(X_train, Y_train, epochs=200, verbose=verbose)
    if get_history:
        return h.history['loss'], model
    else:
        return model


def SAE(X_train,Y_train,init='zero'):
    encoded_X, encoder_0 = model1(X_train)
    encoder_1 = model2(X_train,encoded_X,encoder_0)
    h, model = model3(X_train, Y_train, encoder_0, encoder_1,init, get_history=True)
    return h,model