def setup_and_parse_conf():
parser = ArgumentParser()
parser.add_argument('config', help='Path to configuration file')
parser.add_argument('--output_dir', help='where to store output')
arguments = parser.parse_args()
with open(arguments.config) as f:
config = json.loads(f.read())
_output_dir = conf_dir =\
path.dirname(path.join(os.getcwd(), arguments.config))
if arguments.output_dir:
_output_dir = path.join(os.getcwd(), arguments.output)
training_file = path.join(_output_dir, 'training-dataset.pickle')
test_file = path.join(_output_dir, 'test-dataset.pickle')
if path.exists(training_file) and path.exists(test_file):
_training_data = pickle.load(open(training_file, 'r'))
_test_data = pickle.load(open(test_file, 'r'))
print "Using pickled training and test data from files (%s, %s)" % (
training_file, test_file)
else:
_training_data = create_datasets(**config['datasets']['training'])[0]
_test_data = create_datasets(**config['datasets']['test'])[0]
pickle.dump(_training_data, open(training_file, 'w'))
pickle.dump(_test_data, open(test_file, 'w'))
print "Created fresh training and test data, stored in (%s, %s)" % (
training_file, test_file)
class AnalyticsConfiguration():
output_dir = _output_dir
# System
n_cores = config['system']['n_cores']
# Datasets
training_data = _training_data
test_data = _test_data
# Reservoir
connectivity = config['reservoir']['connectivity']
# Distribution
n_nodes_range = range(*config['distribution']['n_nodes_range'])
n_samples = config['distribution']['n_samples']
def input_connectivity_fn(self, *args):
return eval(config['distribution']['input_connectivity_fn'])(*args)
def output_connectivity_fn(self, *args):
return eval(
config['distribution']['output_connectivity_fn'])(*args)
return AnalyticsConfiguration()
def setup_and_parse_conf():
parser = ArgumentParser()
parser.add_argument('config', help='Path to configuration file')
parser.add_argument('--output_dir', help='where to store output')
arguments = parser.parse_args()
with open(arguments.config) as f:
config = json.loads(f.read())
_output_dir = conf_dir =\
path.dirname(path.join(os.getcwd(), arguments.config))
if arguments.output_dir:
_output_dir = path.join(os.getcwd(), arguments.output)
training_file = path.join(_output_dir, 'training-dataset.pickle')
test_file = path.join(_output_dir, 'test-dataset.pickle')
if path.exists(training_file) and path.exists(test_file):
_training_data = pickle.load(open(training_file, 'r'))
_test_data = pickle.load(open(test_file, 'r'))
print "Using pickled training and test data from files (%s, %s)" % (training_file, test_file)
else:
_training_data = create_datasets(**config['datasets']['training'])[0]
_test_data = create_datasets(**config['datasets']['test'])[0]
pickle.dump(_training_data, open(training_file, 'w'))
pickle.dump(_test_data, open(test_file, 'w'))
print "Created fresh training and test data, stored in (%s, %s)" % (training_file, test_file)
class AnalyticsConfiguration():
output_dir = _output_dir
# System
n_cores = config['system']['n_cores']
# Datasets
training_data = _training_data
test_data = _test_data
# Reservoir
connectivity = config['reservoir']['connectivity']
# Distribution
n_nodes_range = range(*config['distribution']['n_nodes_range'])
n_samples = config['distribution']['n_samples']
def input_connectivity_fn(self, *args):
return eval(config['distribution']['input_connectivity_fn'])(*args)
def output_connectivity_fn(self, *args):
return eval(config['distribution']['output_connectivity_fn'])(*args)
return AnalyticsConfiguration()
def estimate_reservoir_distribution(n_samples, n_nodes, connectivity,
input_connectivity_range, window_size):
results = []
datasets = create_datasets(20,
task_size=200,
window_size=window_size,
dataset_type='temporal_parity')
training_dataset, test_dataset = datasets[:-1], datasets[-1]
for input_connectivity in input_connectivity_range:
logging.info('Sampling N={} with L={}.'.format(n_samples,
input_connectivity))
for sample in range(n_samples):
try:
reservoir = RBNNode(connectivity=connectivity,
output_dim=n_nodes,
input_connectivity=input_connectivity)
readout = Oger.nodes.RidgeRegressionNode(
input_dim=reservoir.output_dim, output_dim=1)
flow = mdp.Flow([reservoir, readout], verbose=1)
flow.train([None, training_dataset])
accuracy = calculate_accuracy(flow, test_dataset)
#cc = measure_computational_capability(reservoir, 100, window_size)
#result = [input_connectivity, accuracy, cc]
results.append([accuracy, reservoir])
#results.append(result)
logging.info(accuracy)
except Exception as e:
logging.error(e)
logging.error('Exception occured, Continuing anyways')
logging.info(results)
return results
def create_dataset():
dataset_type = default_input('Dataset [temporal_parity, temporal_density]',
'temporal_parity')
n_datasets = default_input('Datasets', 10)
task_size = default_input('Dataset length', 200)
window_size = default_input('Window size', 3)
datasets = temporal.create_datasets(n_datasets,
task_size=task_size,
window_size=window_size,
dataset_type=dataset_type)
dataset_description = '[{}-{}-{}-{}]'.format(dataset_type, n_datasets,
task_size, window_size)
return datasets, dataset_description
def create_dataset():
dataset_type = default_input(
'Dataset [temporal_parity, temporal_density]', 'temporal_parity')
n_datasets = default_input('Datasets', 10)
task_size = default_input('Dataset length', 200)
window_size = default_input('Window size', 3)
datasets = temporal.create_datasets(
n_datasets,
task_size=task_size,
window_size=window_size,
dataset_type=dataset_type)
dataset_description = '[{}-{}-{}-{}]'.format(
dataset_type, n_datasets, task_size, window_size)
return datasets, dataset_description
def estimate_reservoir_distribution(n_samples, n_nodes, connectivity,
input_connectivity_range, window_size):
results = []
datasets = create_datasets(
20,
task_size=200,
window_size=window_size,
dataset_type='temporal_parity')
training_dataset, test_dataset = datasets[:-1], datasets[-1]
for input_connectivity in input_connectivity_range:
logging.info('Sampling N={} with L={}.'
.format(n_samples, input_connectivity))
for sample in range(n_samples):
try:
reservoir = RBNNode(connectivity=connectivity,
output_dim=n_nodes,
input_connectivity=input_connectivity)
readout = Oger.nodes.RidgeRegressionNode(
input_dim=reservoir.output_dim,
output_dim=1)
flow = mdp.Flow([reservoir, readout], verbose=1)
flow.train([None, training_dataset])
accuracy = calculate_accuracy(flow, test_dataset)
#cc = measure_computational_capability(reservoir, 100, window_size)
#result = [input_connectivity, accuracy, cc]
results.append([accuracy, reservoir])
#results.append(result)
logging.info(accuracy)
except Exception as e:
logging.error(e)
logging.error('Exception occured, Continuing anyways')
logging.info(results)
return results
from tasks.temporal import create_datasets
from rbn.sklearn_rbn_node import RBNNode
from sklearn.linear_model import Ridge
datasets = create_datasets(10,
task_size=200,
window_size=3,
dataset_type='temporal_parity')
training_datasets, test_dataset = datasets[:-1], datasets[-1]
rbn_reservoir = RBNNode(connectivity=3,
input_connectivity=50,
n_nodes=100,
output_connectivity=0)
readout_layer = Ridge()
for td in training_datasets[:1]:
#rbn_reservoir.reset_state()
reservoir_input, expected_output = td
states = rbn_reservoir.execute(reservoir_input)
readout_layer.fit(states, expected_output)
predictions = readout_layer.predict(states)
for prediction in predictions:
prediction[0] = 1 if prediction[0] > 0.5 else 0
errors = sum(predictions != expected_output)
accuracy = 1 - float(errors) / len(predictions)
from tasks.temporal import create_datasets
from rbn.sklearn_rbn_node import RBNNode
from sklearn.linear_model import Ridge
datasets = create_datasets(
10,
task_size=200,
window_size=3,
dataset_type='temporal_parity')
training_datasets, test_dataset = datasets[:-1], datasets[-1]
rbn_reservoir = RBNNode(connectivity=3,
input_connectivity=50,
n_nodes=100,
output_connectivity=0)
readout_layer = Ridge()
for td in training_datasets[:1]:
rbn_reservoir.reset_state()
reservoir_input, expected_output = td
states = rbn_reservoir.execute(reservoir_input)
readout_layer.fit(states, expected_output)
predictions = readout_layer.predict(states)
for prediction in predictions:
prediction[0] = 1 if prediction[0] > 0.5 else 0
errors = sum(predictions != expected_output)
accuracy = 1 - float(errors) / len(predictions)