def test_estimate(self):
config = {
"n_attributes": 2,
"learning_rate": 0.1,
"batch_size": 1,
}
data_set = DataFrame({'id1': [2, 2, 1, 1, 1, ], 'id2': [1, 1, 2, 2, 2, ], 'rating': [2, 2, 3, 3, 3, ]})
estimator = Estimator(data_set.ix[:, :2].values, config, 2, 2)
y = data_set.ix[:, 2].values
for metric in ['MSE', 'MAE']:
error = estimator.estimate(y, config['batch_size'], 0.1, metric, 8)
assert error > 0
assert error < 1
def evaluate(data_set_name, layer_size, n_hidden_layers):
with open(os.path.join('../specs', data_set_name + '.json')) as specs_file:
specs = json.load(specs_file)
data_set = pandas.read_csv(os.path.join('../resources', specs['file']), sep=specs['separator'],
engine=specs['engine'])
print(data_set.head())
with open(os.path.join(os.path.dirname(__file__), 'neural-net.json')) as config_file:
config = json.load(config_file)
x = data_set.ix[:, :2].values
estimator = Estimator(x, config, layer_size, n_hidden_layers)
y = data_set.ix[:, 2].values.reshape(-1, 1)
if specs['scaling']:
y = sklearn.preprocessing.MaxAbsScaler().fit_transform(numpy.log(y))
return estimator.estimate(y, config['batch_size'], specs['test_size'], specs['metric'])
def evaluate(data_set_name, layer_size, n_hidden_layers):
with open(os.path.join('../specs', data_set_name + '.json')) as specs_file:
specs = json.load(specs_file)
data_set = pandas.read_csv(os.path.join('../resources', specs['file']), sep=specs['separator'],
engine=specs['engine'])
print(data_set.head())
with open(os.path.join(os.path.dirname(__file__), 'neural-net.json')) as config_file:
config = json.load(config_file)
x = data_set.ix[:, :2].values
estimator = Estimator(x, config, layer_size, n_hidden_layers)
y = data_set.ix[:, 2].values.reshape(-1, 1)
if specs['scaling']:
y = sklearn.preprocessing.MaxAbsScaler().fit_transform(numpy.log(y))
return estimator.estimate(y, config['batch_size'], specs['test_size'], specs['metric'])
