model.add(Dense(256, activation='relu'))
model.add(Dropout(dropout))
# Dense layer 3/sigmoid boi
model.add(Dense(1, activation='sigmoid'))
# Compile model
optimizer = optimizers.Adam(lr=learning_rate, decay=0.0)
model.compile(loss='binary_crossentropy',
optimizer=optimizer,
metrics=['accuracy'])
model.summary()
return model
if __name__ == '__main__':
X, y, _ = get_input_data(train_file_path='train.json')
# Incorporate rotated images into training data (note that this significantly increases training time)
X_rotated, y_rotated = get_rotated_images(X, y)
X = np.concatenate([X, X_rotated])
y = np.concatenate([y, y_rotated])
X_train = X
y_train = y
X_test, ids = get_input_data('data/test.json', train=False)
model = get_model(learning_rate=0.001, dropout=0.2)
# Train and test model
train_history = model.fit(X_train,
y_train,
import numpy as np
from numpy.random import seed
seed(7)
from tensorflow import set_random_seed
set_random_seed(420)
import csv
import keras
from data_preprocessing import get_input_data
from sklearn.model_selection import train_test_split
import os
from sklearn.metrics import accuracy_score, precision_score, recall_score
if __name__ == '__main__':
X, y = get_input_data(train_file_path='train.json')
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.3,
random_state=42)
saved_models = [f for f in os.listdir('.') if f.startswith('hyperparams_')]
with open('hyperparamater_search_results.csv', 'w') as csv_file:
writer = csv.writer(csv_file)
header = [
'learning_rate', 'epochs', 'batch_size', 'dropout', 'accuracy',
'log_loss', 'precision', 'recall'
]
writer.writerow(header)