def test_mlp_whale(self):
X_train, Y_train, X_test, Y_test = data.load_whale_data(
WHALE_TRAIN_DATA, WHALE_TEST_DATA, dim=1)
input_shape = (X_train.shape[1], )
nb_classes = 447
for nb_neurons in [447, 500, 600, 800, 1000]:
# for nb_neurons_2 in [447, 500, 600, 800, 1000]:
# for nb_neurons_3 in [447, 500, 600, 800, 1000]:
# target_path = "mlp_{0}_{1}_{2}_447".format(
# nb_neurons,
# nb_neurons_2,
# nb_neurons_3
# )
target_path = "mlp_{0}_447".format(nb_neurons)
mkdir_p(target_path)
mlp.mlp(X_train=X_train,
Y_train=Y_train,
X_test=X_test,
Y_test=Y_test,
input_shape=input_shape,
nb_classes=nb_classes,
nb_layers=2,
hidden_neurons=[nb_neurons, 447],
activations=["tanh", "tanh", "tanh", "tanh"],
dropouts=[0.1, 0.1, 0.1, 0.1, 0.1],
loss='categorical_crossentropy',
optimizer="adadelta",
nb_epoch=30,
batch_size=100,
model_file=os.path.join(target_path, "model.json"),
results_file=os.path.join(target_path, "results.dat"),
weights_file=os.path.join(target_path, "weights.dat"))
def test_cnn_whale(self):
X_train, Y_train, X_test, Y_test = data.load_whale_data(
WHALE_TRAIN_DATA, WHALE_TEST_DATA)
# X_train = []
# Y_train = []
# X_test = []
# Y_test = []
kwargs = {
"X_train": X_train,
"Y_train": Y_train,
"X_test": X_test,
"Y_test": Y_test,
"input_shape": (1, 192, 192),
"nb_classes": 477,
"nb_convo_layers": 2,
"nb_filters": [32, 32],
"nb_conv": [3, 3],
"convo_activations": ["relu", "relu"],
"maxpools": [False, True],
"pool_sizes": [None, 2],
"convo_dropouts": [None, 0.25],
"nb_dense_layers": 1,
"dense_hidden_neurons": [477],
"dense_activations": ["relu"],
"dense_dropouts": [0.5],
"loss": "categorical_crossentropy",
"optimizer": "adadelta",
"nb_epoch": 30,
"batch_size": 32,
"weights_file": "weights.hdf5",
"results_file": "results.dat"
}
fitlog, score = cnn.cnn(**kwargs)
def test_run(self):
random.seed(42)
X_train, Y_train, X_test, Y_test = data.load_whale_data(
WHALE_TRAIN_DATA,
WHALE_TEST_DATA
)
# neural network data
nn_data = {
"X_train": X_train,
"Y_train": Y_train,
"X_test": X_test,
"Y_test": Y_test,
"input_shape": (1, 96, 96),
"n_outputs": 447,
"model_save_dir": "/data/nn_exp"
}
# run
ga.run(
nn_data=nn_data,
chromo_size=500,
max_gen=20,
pop_size=20,
t_size=2,
record_file_path="execution_test.dat",
score_file_path="score_test.dat",
error_file_path="error_test.dat"
)
def test_denoising_autoencoder(self):
# # load data
X_train, Y_train, X_test, Y_test = data.load_whale_data(
WHALE_TRAIN_DATA, WHALE_TEST_DATA, dim=1)
# autoencoder parameters
batch_size = 32
nb_epoch = 1000
nb_neurons = 600
img_dim = X_train.shape[1]
# create and train
X_train_tmp = np.copy(X_train)
ae = Sequential()
ae.add(
AutoEncoder(
encoder=Dense(nb_neurons,
input_dim=img_dim,
activation='sigmoid'),
decoder=Dense(img_dim,
input_dim=nb_neurons,
activation='sigmoid'),
output_reconstruction=True,
))
# compile
ae.compile(loss="mean_squared_error", optimizer="adadelta")
# fit
ae.fit(X_train_tmp,
X_train_tmp,
batch_size=batch_size,
nb_epoch=nb_epoch)
results = ae.predict(X_train, verbose=1)
np.save("results_2.npy", results)
img_data = np.concatenate((X_train[0:9], results[0:9]))
data.plot_multiple_mnist_images(img_data[0:9], (6, 3))
#!/usr/bin/env python2
import os
import sys
import random
sys.path.append(os.path.join(os.path.dirname(__file__), "../../"))
import recognizer.data as data
import recognizer.tuner.ga as ga
# GLOBAL VARIABLES
WHALE_TRAIN_DATA = "/data/whale_data/manual_rotated/train_data.csv"
WHALE_TEST_DATA = "/data/whale_data/manual_rotated/test_data.csv"
if __name__ == "__main__":
random.seed(42)
X_train, Y_train, X_test, Y_test = data.load_whale_data(
WHALE_TRAIN_DATA, WHALE_TEST_DATA)
# neural network data
train_end = len(X_train) / 0.5
test_end = len(X_test) / 0.5
nn_data = {
"X_train": X_train[0:train_end],
"Y_train": Y_train[0:train_end],
"X_test": X_test[0:test_end],
"Y_test": Y_test[0:test_end],
"input_shape": (1, 96, 96),
"n_outputs": 447,
"model_save_dir": "/data/ga_cnn_dataset_3"
}
# run
import recognizer.data as data
import recognizer.tuner.ga as ga
# GLOBAL VARIABLE
WHALE_TRAIN_DATA = "/home/chutsu/whale_data/train.csv"
WHALE_TEST_DATA = "/home/chutsu/whale_data/test.csv"
if __name__ == "__main__":
seed = int(sys.argv[1])
random.seed(42 + seed)
# neural network data
X_train, Y_train, X_test, Y_test = data.load_whale_data(
WHALE_TRAIN_DATA,
WHALE_TEST_DATA
)
train_end = int(round(len(X_train) * 0.01))
test_end = int(round(len(X_test) * 0.01))
nn_data = {
"X_train": X_train[:train_end],
"Y_train": Y_train[:train_end],
"X_test": X_test[:test_end],
"Y_test": Y_test[:test_end],
"input_shape": (1, 192, 192),
"n_outputs": 447
}
# run
ga.run(
nn_data=nn_data,