import tensorflow as tf from tensorflow.keras.models import Sequential, Model from tensorflow.keras.layers import Convolution2D, MaxPooling2D, Dropout, Flatten, LSTM

Step 3: Initialize

def Initialize(): # code for initialization pass

Step 4: Normalization

def Normalization(Train_X, Train_Y): # code for normalization pass

Step 5: Define Convolution_1 model

def Convolution_1(optimizer, dropout, poolsize): model = Sequential() model.add(Convolution2D(optimizer, dropout, name="Conv2D_1")) model.add(MaxPooling2D(poolsize)) model.add(Dropout(dropout)) return model

Step 10: Define Lstmmodel

def Lstmmodel(units, activation, recurrent_activation): model = Sequential() model.add(LSTM(units, activation=activation, recurrent_activation=recurrent_activation)) model.add(Flatten()) return model

Step 12: Compile and fit Lstmmodel

def compile_and_fit(model, lossfunction, optimizer, Convolution_1_feature, Train_Y, batchsize, epochs): model.compile(loss=lossfunction, optimizer=optimizer) model.fit(Convolution_1_feature, Train_Y, batch_size=batchsize, epochs=epochs)

Step 1: Input: Train_X, Train_Y

Train_X = ... Train_Y = ...

Step 2: Hyper-Parameters

optimizer = ... rate = ... feature_layers = ... poolsize = ... batchsize = ...

Step 3: Initialize

Initialize()

Step 4: Normalization

Normalization(Train_X, Train_Y)

Step 5: Define Convolution_1 model

Convolution_1_model = Convolution_1(optimizer, rate, poolsize)

Step 6: Compile Convolution_1 model

Convolution_1_model.compile(optimizer, loss='categorical_crossentropy')

Step 7: Fit Convolution_1 model

Convolution_1_model.fit(Train_X, Train_Y, epochs=epochs, batch_size=batchsize)

Step 8: Define Convolution_1_feature model

Convolution_1_feature_model = Model(inputs=Convolution_1_model.input, outputs=Convolution_1_model.get_layer('Conv2D_1').output)

Step 9: Predict Convolution_1_feature

Convolution_1_feature = Convolution_1_feature_model.predict(Train_X)

Step 10: Define Lstmmodel

Lstmmodel_model = Lstmmodel(units, activation, recurrent_activation)

Step 11: Compile and fit Lstmmodel

compile_and_fit(Lstmmodel_model, lossfunction, optimizer, Convolution_1_feature, Train_Y, batchsize, epochs)