# summarize results
#print("Best: %f using %s" % (grid_result.best_score_, grid_result.best_params_))
# create model
model = build_model(x_train.shape[1:], y_train.shape[-1], activation[0],
learn_rate[0], loss[0], optimizer[0], drop_rate[0])
print('built model..')
# Data Augmentation
datagen = ImageDataGenerator(featurewise_center=True,
featurewise_std_normalization=True,
rotation_range=0.0,
fill_mode='nearest',
horizontal_flip=True,
vertical_flip=True,
rescale=1. / 255,
preprocessing_function=None,
validation_split=0.25)
datagen.fit(x_train)
model.fit_generator(datagen.flow(x_train, y_train, batch_size=batch_size[0]),
epochs=epochs[0])
# Save model with weights
model.save(data_dir + '/models/yelp_model.h5')
score = model.evaluate(x_test, y_test)
print('Model loss:', score)
#print('Model accuracy:', score)
ltk_keras = ltk
model = KerasClassifier(build_fn= ltk_keras.Build_Model, verbose=1)
hyperparameters = ltk_keras.create_hyperparameters()
search = RandomizedSearchCV(estimator=model,
param_distributions=hyperparameters,
n_iter=10, n_jobs=-1, cv=3, verbose=1)
Max_acc = 0
data_generator = ImageDataGenerator(
rotation_range=20,
width_shift_range=0.02,
height_shift_range=0.02,
horizontal_flip=True
)
for i in range(10):
search.fit(X_list_train[i], Y_list_train[i])
print(search.best_params_)
model = ltk_keras.Build_Model(search.best_params_["keep_prob"], search.best_params_['optimizer'], search.best_params_['kernel_regularizer_num'])
history = model.fit_generator(
data_generator.flow(X_list_train[i], Y_list_train[i], batch_size=search.best_params_["batch_size"]),
steps_per_epoch= IMG_ROWS * IMG_COLS * IMG_CHANNELS,
epochs = 10,
validation_data = (X_test, Y_test),
verbose=1
)
loss, acc = model.evaluate(X_test, Y_test)
print("정답률: ", acc)
if Max_acc < acc:
Max_acc = acc
print("최종정답률: ", acc)
grid_result = grid.fit(base_model, outputs)
# summarize results
print("Best: %f using %s" % (grid_result.best_score_, grid_result.best_params_))
for params, mean_score, scores in grid_result.grid_scores_:
print("%f (%f) with: %r" % (scores.mean(), scores.std(), params))
early_stopping = EarlyStopping(patience=20)
checkpointer = ModelCheckpoint('inception_resnet_bottleneck_drug_best.h5', verbose=1, save_best_only=True)
ImageFile.LOAD_TRUNCATED_IMAGES = True
model.fit_generator(batches, steps_per_epoch=num_train_steps, epochs=1000, callbacks=[early_stopping, checkpointer], validation_data=val_batches, validation_steps=num_valid_steps)
model.save_weights('inception_resnet_bottleneck_drug_weights.h5')
model.save('inception_resnet_bottleneck_drug.h5')
# for layer in model.layers[-31:]:
# layer.trainable=True
# for layer in model.layers[:-31]:
# layer.trainable=False
# model.compile(loss='categorical_crossentropy', optimizer=optimizers.SGD(lr=1e-4, momentum=0.9), metrics=['accuracy'])
# checkpointer = ModelCheckpoint('./resnet50_best_safety.h5', verbose=1, save_best_only=True)
# model.fit_generator(batches, steps_per_epoch=num_train_steps, epochs=1000, callbacks=[early_stopping, checkpointer], validation_data=val_batches, validation_steps=num_valid_steps)
# model.save('resnet50_safety.h5')
def train_model(param):
train_out = np.load('data/train_out.npy')
test_out = np.load('data/test_out.npy')
train_norm_2d = np.load('data/train_norm_2d.npy')
test_norm_2d = np.load('data/test_norm_2d.npy')
train_norm_2d_new = np.load('data/train_norm_2d_new.npy')
train_out_new = np.load('data/train_out_new.npy')
out_threshold = 0.005
####################### GRID HYPERPARAMETERS ###########################
KFolds = 3
e = [5, 10, 15]
learning_rate = [0.01, 0.005, 0.001]
param_grid = dict(epochs=e, learning_rate=learning_rate)
####################### CV GRID SEARCH #################################
augmentation = False
t0 = time.time()
l = []
model = KerasClassifier(build_fn=create_model, batch_size=param)
if augmentation == True:
datagen = ImageDataGenerator(rotation_range=5,
width_shift_range=0,
height_shift_range=0,
shear_range=0,
zoom_range=0,
horizontal_flip=True,
fill_mode='nearest')
#datagen.fit(train_norm_2d_new)
print(train_norm_2d_new.shape)
print(train_out_new.shape)
print("Running augmented training now, with augmentation")
history = model.fit_generator(datagen.flow(train_norm_2d_new,
train_out_new,
batch_size=param,
shuffle=True),
steps_per_epoch=len(train_norm_2d_new) /
param,
epochs=10)
indices_test = np.where(test_out > out_threshold)[0]
test_out_pos = test_out[indices_test]
test_out_new = np.tile(test_out_pos, 18)
test_out_new = np.concatenate((test_out, test_out_new), axis=0)
test_norm_2d_pos = test_norm_2d[indices_test, :, :, :]
test_norm_2d_new = np.tile(test_norm_2d_pos, (18, 1, 1, 1))
test_norm_2d_new = np.concatenate((test_norm_2d, test_norm_2d_new),
axis=0)
t1 = time.time()
return calc_verification_scores(model, test_norm_2d_new,
test_out_new), t1 - t0
else:
print("Running regular training, no augmentation")
#ros=RandomOverSampler(random_state=0)
#train_norm_2d_resampled, train_out_resampled=ros.fit_resample(train_norm_2d,train_out)
start_time = time.time()
grid = GridSearchCV(estimator=model,
param_grid=param_grid,
cv=KFolds,
n_jobs=-1)
grid_result = grid.fit(train_norm_2d_new, train_out_new)
means = grid_result.cv_results_['mean_test_score']
stds = grid_result.cv_results_['std_test_score']
params = grid_result.cv_results_['params']
end_time = time.time()
run_time = end_time - start_time
#preds = model.predict(test_norm_2d)
for mean, stdev, p in zip(means, stds, params):
x = [("%f (%f) with: %r, batch: %r ,time: %r" %
(mean, stdev, p, param, run_time))]
l.append(x)
#calc_verification_scores(test_out,preds)
return l
combined_pred = test_crs.merge(pred_cols, left_index=True, right_index=True)
combined_pred.shape
combined_pred.to_csv('/home/joseh/data/cnn_cadr_student_testset_title.csv',
encoding='utf-8',
index=False)
## Try k-fold cross validation
for j, (train_idx, val_idx) in enumerate(folds):
print('\nFold ', j)
X_train_cv = X_train[train_idx]
y_train_cv = y_train[train_idx]
X_valid_cv = X_train[val_idx]
y_valid_cv = y_train[val_idx]
name_weights = "final_model_fold" + str(j) + "_weights.h5"
callbacks = get_callbacks(name_weights=name_weights, patience_lr=10)
generator = gen.flow(X_train_cv, y_train_cv, batch_size=batch_size)
model = get_model()
model.fit_generator(generator,
steps_per_epoch=len(X_train_cv) / batch_size,
epochs=15,
shuffle=True,
verbose=1,
validation_data=(X_valid_cv, y_valid_cv),
callbacks=callbacks)
print(model.evaluate(X_valid_cv, y_valid_cv))
