I am currently struggling with getting good results with my CNN in which i
decided to optimize parameter using grid search. I am currently trying to use
scikit-learn GridSearchCV.
def
create_model(init_mode='uniform',activation_mode='linear',optimizer_mode="adam",
activation_mode_conv = 'linear'):
model = Sequential()
model.add(ZeroPadding2D((6,4),input_shape=(6,3,3)))
model.add(Convolution2D(32,3,3 , activation=activation_mode_conv))
print model.output_shape
model.add(Convolution2D(32, 3,3, activation=activation_mode_conv))
print model.output_shape
model.add(MaxPooling2D(pool_size=(2,2),strides=(2,1)))
print model.output_shape
model.add(Convolution2D(64, 3,3 , activation=activation_mode_conv))
print model.output_shape
model.add(Convolution2D(64, 3,3 , activation=activation_mode_conv))
print model.output_shape
model.add(MaxPooling2D(pool_size=(2,2),strides=(2,1)))
model.add(Flatten())
print model.output_shape
model.add(Dense(output_dim=32, input_dim=64,
init=init_mode,activation=activation_mode))
model.add(Dense(output_dim=13, input_dim=50,
init=init_mode,activation=activation_mode))
model.add(Dense(output_dim=1, input_dim=13,
init=init_mode,activation=activation_mode))
model.add(Dense(output_dim=1, init=init_mode,
activation=activation_mode))
#print model.summary()
model.compile(loss='mean_squared_error',optimizer=optimizer_mode)
return model
#reduce_lr=ReduceLROnPlateau(monitor='val_loss', factor=0.01,
patience=3, verbose=1, mode='auto', epsilon=0.1, cooldown=0,
min_lr=0.000000000000000001)
#stop = EarlyStopping(monitor='val_loss', min_delta=0, patience=5,
verbose=1, mode='auto')
#log=csv_logger = CSVLogger('training_'+str(i)+'.csv')
#print "Model Train"
#hist_current = model.fit(np.array(data_train_input),
# np.array(data_train_output),
# shuffle=False,
#
validation_data=(np.array(data_test_input),np.array(data_test_output)),
# validation_split=0.1,
# nb_epoch=150000,
# verbose=1,
# callbacks=[reduce_lr,log,stop])
#print()
#print model.summary()
#print "Model stored"
#model.save(spectogram_path+"Model"+str(feature)+".h5")
#model.save_weights(spectogram_path+"Model"+str(feature)+"_weights.h5")
#del model
## Make it work for other feature ranges
## Add the CNN part and test it
## Try with gabor kernels as suggested by the other paper..
input_train, input_test, output_train, output_test =
model(0,train_input_data_interweawed_normalized[:-(len(train_input_data_interweawed_normalized)-1000)],output_data_train[:-(len(output_data_train)-1000)],test_input_data_interweawed_normalized[:-(len(test_input_data_interweawed_normalized)-1000)],output_data_test[:-(len(output_data_test)-1000)])
del test_input_data
del test_name
del test_input_data_normalized
del test_name_normalized
del test_input_data_interweawed
del test_name_interweawed
del test_input_data_interweawed_normalized
del test_name_interweawed_normalized
del train_input_data
del train_name
del train_input_data_normalized
del train_name_normalized
del train_input_data_interweawed
del train_name_interweawed
del train_input_data_interweawed_normalized
del train_name_interweawed_normalized
seed = 7
np.random.seed(seed)
print "Regressor"
model = KerasRegressor(build_fn = create_model, verbose = 10)
init_mode_list = ['uniform', 'lecun_uniform', 'normal', 'zero',
'glorot_normal', 'glorot_uniform', 'he_normal', 'he_uniform']
activation_mode_list = ['softmax', 'softplus', 'softsign', 'relu', 'tanh',
'sigmoid', 'hard_sigmoid', 'linear']
activation_mode_list_conv = ['softplus', 'softsign', 'relu', 'tanh',
'sigmoid', 'hard_sigmoid', 'linear']
optimizer_mode_list = ['SGD', 'RMSprop', 'Adagrad', 'Adadelta', 'Adam',
'Adamax', 'Nadam']
batch_size_list = [10, 20, 40, 60, 80, 100]
epochs = [10, 50, 100]
param_grid = dict(init_mode=init_mode_list, batch_size=batch_size_list,
nb_epoch=epochs, activation_mode=activation_mode_list, optimizer_mode =
optimizer_mode_list, activation_mode_conv = activation_mode_list_conv)
grid = GridSearchCV(estimator=model, param_grid=param_grid, n_jobs=1)
print "Grid fit"
grid_result = grid.fit(np.asarray(input_train), np.array(output_train))
# summarize results
print("Best: %f using %s" % (grid_result.best_score_,
grid_result.best_params_))
means = grid_result.cv_results_['mean_test_score']
stds = grid_result.cv_results_['std_test_score']
params = grid_result.cv_results_['params']
for mean, stdev, param in zip(means, stds, params):
print("%f (%f) with: %r" % (mean, stdev, param))
This runs.. but problems with it that it only provides a result at the end. I
ran the code once but then it crashed with this error message:
cannot allocate memory for thread-local data: ABORT
I am not sure what could cause this problem?
_______________________________________________
scikit-learn mailing list
scikit-learn@python.org
https://mail.python.org/mailman/listinfo/scikit-learn
