def test_get_error_unusual_targets_shape():
from learning import error
model = multioutputs.MultiOutputs([
helpers.SetOutputModel([1.0]),
helpers.SetOutputModel([1.0, 1.0, 1.0])
])
assert validation.get_error(model, [[]], [[[1.0], [1.0, 1.0, 1.0]]],
error_func=MeanSquaredError()) == 0.0
assert validation.get_error(model, [[]], [[[1.0], [0.0, 0.0, 0.0]]],
error_func=MeanSquaredError()) == 0.5
def test_benchmark(monkeypatch):
# Patch time.clock so time attribute is deterministic
monkeypatch.setattr(time, 'clock', lambda : 0.0)
# Make network that returns set output for
patterns = [
([0], [1]),
([1], [1]),
([2], [1])
]
model = helpers.SetOutputModel([1])
# Track patterns for training
training_patterns = []
def post_pattern_callback(network_, input_vec, target_vec):
training_patterns.append((list(input_vec), list(target_vec)))
# Cross validate with deterministic network, and check output
stats = validation.benchmark(model, zip(*patterns), num_folds=3, num_runs=2,
iterations=1,
post_pattern_callback=post_pattern_callback)
# Check
assert (helpers.fix_numpy_array_equality(stats)
== helpers.fix_numpy_array_equality(_BENCHMARK_STATS))
assert training_patterns == [([1], [1]), ([2], [1]), # First fold
([0], [1]), ([2], [1]), # Second fold
([0], [1]), ([1], [1]), # Third fold
([1], [1]), ([2], [1]), # First fold 2
([0], [1]), ([2], [1]), # Second fold 2
([0], [1]), ([1], [1])] # Third fold 2
def test_get_error():
model = helpers.SetOutputModel([1])
assert validation.get_error(
model,
numpy.array([[1]]),
numpy.array([[0]]),
error_func=MeanSquaredError()) == 1.0
assert validation.get_error(
model,
numpy.array([[1]]),
numpy.array([[1]]),
error_func=MeanSquaredError()) == 0.0
assert validation.get_error(
model,
numpy.array([[1]]),
numpy.array([[0.5]]),
error_func=MeanSquaredError()) == 0.25
assert validation.get_error(
model,
numpy.array([[1], [1]]),
numpy.array([[1], [0]]),
error_func=MeanSquaredError()) == 0.5
assert validation.get_error(
model,
numpy.array([[1], [1]]),
numpy.array([[0.5], [0.5]]),
error_func=MeanSquaredError()) == 0.25
def test_get_accuracy():
model = helpers.SetOutputModel([1])
assert validation.get_accuracy(model, numpy.array([[1], [1]]),
numpy.array([[1], [0]])) == 0.5
assert validation.get_accuracy(model, numpy.array([[1], [1]]),
numpy.array([[1], [1]])) == 1.0
assert validation.get_accuracy(model, numpy.array([[1], [1]]),
numpy.array([[0], [0]])) == 0.0
def test_break_on_no_improvement_completely_stagnant():
nn = helpers.SetOutputModel(1.0)
# Stop training if error does not improve after 5 iterations
nn.train([[0.0]], [[0.0]],
error_stagnant_distance=10,
error_stagnant_threshold=None,
error_improve_iters=5)
assert nn.iteration == 6 # The 6th is 5 away from the first
def test_bagger():
# Create dummy layers that return set outputs
outputs = [[0, 1, 2], [1, 2, 3]]
models = [helpers.SetOutputModel(output) for output in outputs]
bagger = ensemble.Bagger(models)
# Assert bagger returns average of those outputs
output = bagger.activate([])
assert list(output) == [0.5, 1.5, 2.5]
def test_compare(monkeypatch):
# Patch time.clock so time attribute is deterministic
monkeypatch.setattr(time, 'clock', lambda: 0.0)
# Make network that returns set output for
patterns = [([0], [1]), ([1], [1]), ([2], [1])]
model = helpers.SetOutputModel([1])
model2 = helpers.SetOutputModel([1])
# Cross validate with deterministic network, and check output
stats = validation.compare(['model', 'model2'], [model, model2],
zip(*patterns),
num_folds=3,
num_runs=2,
all_kwargs={'iterations': 1})
# Check
assert (helpers.fix_numpy_array_equality(_drop_models_model_stat(stats)) ==
helpers.fix_numpy_array_equality(_COMPARE_STATS))
def test_break_on_stagnation_completely_stagnant():
# If error doesn't change by enough after enough iterations
# stop training
nn = helpers.SetOutputModel(1.0)
# Stop training if error does not change by more than threshold after
# distance iterations
nn.train([[0.0]], [[0.0]],
error_stagnant_distance=5,
error_stagnant_threshold=0.01)
assert nn.iteration == 6 # The 6th is 5 away from the first
def test_unserialize_wrong_type():
"""Model.unserialize should raise error if serialized model is of wrong type."""
with pytest.raises(ValueError):
base.Model.unserialize(helpers.SetOutputModel(1.0).serialize())
def test_unserialize():
model = helpers.SetOutputModel(random.uniform(0, 1))
model_copy = helpers.SetOutputModel.unserialize(model.serialize())
assert model_copy.__dict__ == model.__dict__, 'Should have same content'
assert model_copy is not model, 'Should have different id'
def test_serialize():
model = helpers.SetOutputModel(1.0)
assert isinstance(model.serialize(),
str), 'Model.serialize should return string'
def test_multioutputs_activate():
model = multioutputs.MultiOutputs(helpers.SetOutputModel(1), 2)
assert model.activate([None]) == [1, 1]
