def test_mnist(self):
from loader import mnist_loader
training_data, validation_data, test_data = \
mnist_loader.load_data_wrapper()
net = network.Network([784, 30, 10])
net.SGD(training_data, 30, 10, 3.0)
self.assertTrue(net.evaluate(test_data)*1.0/len(test_data) >= 0.9)
def test_mnist(self):
from loader import mnist_loader
training_data, validation_data, test_data = \
mnist_loader.load_data_wrapper()
net = network.Network([784, 30, 10])
net.SGD(training_data, 30, 10, 3.0)
self.assertTrue(net.evaluate(test_data) * 1.0 / len(test_data) >= 0.9)
def test_calc_prediction_accuracy():
loaded_weights = load(filename='test_once_sigmoid_alpha_0.04_batch_100_draw_range_0.001_hidden_neurones_50_'
'test_accuracy_0.968.pkl')['weights']
_, _, test_data = load_data_wrapper("../data")
test_data = add_bias(data=test_data)
te_in, te_out = test_data
weights_tested = loaded_weights[0]
assert(calc_prediction_accuracy(SIGMOID.activation, *weights_tested, te_in, te_out) == 0.101)
weights_tested = loaded_weights[-1]
assert(calc_prediction_accuracy(SIGMOID.activation, *weights_tested, te_in, te_out) == 0.9687)
def test_mlp():
loaded_data = load_data_wrapper("../data")
loaded_result = load(filename='test_once_sigmoid_alpha_0.05_batch_100_draw_range_0.2_hidden_neurones_15_mom_0.5_'
'accuracy_0.7321.pkl')
np.random.seed(0)
result = mlp(data=loaded_data, activation=SIGMOID, alpha=0.05, batch_size=100, draw_range=0.2,
hidden_neurones=15, worse_result_limit=1, momentum_param=0.5, use_adagrad=False, images_len_divider=250)
for (w11 , w12), (w21, w22) in zip(loaded_result['weights'], result['weights']):
assert(np.allclose(a=w11, b=w21))
assert(np.allclose(a=w12, b=w22))
assert (len(loaded_result['weights']) == len(result['weights']))
def run_once(name, params, **kwargs):
_, alpha, batch_size, draw_range, hidden_neurones, worse_result_limit, momentum_param, use_adagrad, draw_type \
= (params[p] for p in params)
result = mlp(**params, **kwargs)
print(result)
save(
data=result,
filename=
f'{name}_once_{params["activation"].activation.__name__}_alpha_{alpha}_'
f'batch_{batch_size}_draw_range_{draw_range}_hidden_neurones_{hidden_neurones}_'
f'mom_{momentum_param}_accuracy_{result["test_accuracies"][-1 - worse_result_limit]}.pkl'
)
if __name__ == "__main__":
loaded_data = load_data_wrapper("../data")
run_once(name='',
data=loaded_data,
params=get_simul_params(activation=RELU,
alpha=0.01,
batch_size=10,
draw_range=0.0001,
hidden_neurones=25,
worse_result_limit=3,
momentum_param=0.5))
run_once(name='xavier',
data=loaded_data,
params=get_simul_params(activation=SIGMOID,
alpha=0.03,
batch_size=100,
def train_model():
training_data, validation_data, test_data = \
mnist_loader.load_data_wrapper()
net = network.Network([784, 30, 10])
net.SGD(training_data, 30, 10, 3.0, test_data=test_data)
return net
data['activation'] = 'sigmoid'
str_to_find = '_simulation_'
index = filename.find(str_to_find) + len(str_to_find)
new_filename = f'{filename[:index]}sigmoid_{filename[index:]}'
# data = load(filename)
# hidden_neurones = 'hidden_neurones'
# data[hidden_neurones] = 50
# str_to_find = '_draw_range_'
# index = filename.find(str_to_find) + len(str_to_find) + 4
# new_filename = f'{filename[:index]}hidden_neurones_50_{filename[index:]}'
# save(data=data, filename=new_filename)
print("'" + new_filename + "',")
if __name__ == "__main__":
_, validation_data, test_data = load_data_wrapper("../data")
# print_result('test_weights.pkl', test_data)
# filename = 'test_alpha_0.04_batch_100_draw_range_1.0_hidden_neurones_50.pkl'
# print_result(filename=filename, test_data=test_data)
# file = '2_test_alpha_0.01_batch_100_draw_range_0.05_hidden_neurones_25.pkl'
# file = 'once_sigmoid_alpha_0.04_batch_100_draw_range_0.2_hidden_neurones_100_res_0.9739.pkl'
# test_data = test_data[0], test_data[1]
# print_result(filename=file, test_data=test_data)
# from saver.saver import save
files = [
# 'draw_range_simulation_alpha_0.04_batch_100_draw_range_0.2_hidden_neurones_50_avg_epochs_24.4_times_5.pkl',
# 'draw_range_simulation_alpha_0.04_batch_100_draw_range_0.4_hidden_neurones_50_avg_epochs_24.2_times_5.pkl',
# 'draw_range_simulation_alpha_0.04_batch_100_draw_range_0.6_hidden_neurones_50_avg_epochs_25.0_times_5.pkl',
def train_model():
training_data, validation_data, test_data = \
mnist_loader.load_data_wrapper()
net = network.Network([784, 30, 10])
net.SGD(training_data, 30, 10, 3.0, test_data=test_data)
return net