def __init__(self, *args, **kwargs):
""" Constructor. """
super(NNVisualiseTestCase, self).__init__(*args, **kwargs)
self.cnns = generate_cnn_architectures()
self.mlps_reg = generate_mlp_architectures('reg')
self.mlps_class = generate_mlp_architectures('class')
self.save_dir = '../scratch/unittest_visualisations/'
self.save_dir_for_pres = '../scratch/unittest_visualisations_for_pres/'
def test_mlp_distance_computation(self):
""" Tests the computation of the distance for CNNs. """
self.report('Testing distance computation for specific mlps.')
# Create the transport computation object
mlp_layer_labels, label_mismatch_penalty = \
nn_comparators.get_mlp_layer_label_mismatch_penalties(self.non_assignment_penalty,
'reg')
mlp_tp_comp = nn_comparators.OTMANNDistanceComputer(
mlp_layer_labels,
label_mismatch_penalty,
self.non_assignment_penalty,
nn_comparators.MLP_STRUCTURAL_PENALTY_GROUPS,
nn_comparators.PATH_LENGTH_TYPES,
dflt_mislabel_coeffs=1.0,
dflt_struct_coeffs=1.0)
# Create the mlp architectures
mlps = generate_mlp_architectures()
struct_coeffs = [0.1, 0.2]
dist_types = ['lp-emd', 'lp', 'emd']
mislabel_coeffs = [1.0] * len(struct_coeffs)
self.report(
'Testing distance computation for specific mlps with default coeffs.'
)
self._test_dist_comp_for_single_conn_coeff(mlps, dist_types,
mlp_tp_comp)
self.report(
'Testing distance computation for specific mlps with conn_coeffs=%s.'
% (struct_coeffs))
self._test_dist_comp_for_multiple_coeffs(mlps, dist_types,
mislabel_coeffs,
struct_coeffs, mlp_tp_comp)
def test_call_with_list(self):
""" Tests the __call__ function with a single input of the modifier. """
self.report('Testing the __call__ function with a list of inputs.')
num_modifications = 40
num_steps_probs = [0.5, 0.25, 0.125, 0.075, 0.05]
save_dir = os.path.join(self.save_dir, 'modifier_call_list')
if os.path.exists(save_dir):
rmtree(save_dir)
test_probs = [
self.cnns, self.mlps,
generate_mlp_architectures('class')
]
for idx, prob in enumerate(test_probs):
save_prefix = str(idx)
modifier = self.modifier_wo_cc
modifications = modifier(prob, num_modifications, num_steps_probs)
for new_idx, new_nn in enumerate(modifications):
assert isinstance(new_nn, NeuralNetwork)
visualise_nn(
new_nn,
os.path.join(save_dir, '%s_%d' % (save_prefix, new_idx)))
self.report(
'With list of %d nns(%s):: #new-networks: %d.' %
(len(prob), prob[0].nn_class, len(modifications)),
'test_result')
def __init__(self, *args, **kwargs):
""" Constructor. """
super(NNConstraintCheckerTestCase, self).__init__(*args, **kwargs)
self.nns = generate_cnn_architectures() + generate_mlp_architectures()
self.cnn_constraint_checker = nn_constraint_checkers.CNNConstraintChecker(
25, 5, 500000, 0, 5, 2, 15, 512, 16, 4)
self.mlp_constraint_checker = nn_constraint_checkers.MLPConstraintChecker(
25, 5, 500000, 900, 5, 2, 15, 30, 8)
def __init__(self, *args, **kwargs):
""" Constructor. """
super(NNModifierTestCase, self).__init__(*args, **kwargs)
self.cnns = generate_cnn_architectures()
self.mlps = generate_mlp_architectures()
self.save_dir = '../scratch/unittest_modifier_class/'
self.cnn_constraint_checker = nn_constraint_checkers.CNNConstraintChecker(
50, 4, np.inf, 4.0, 5, 5, 100, 8000, 8)
self.mlp_constraint_checker = nn_constraint_checkers.MLPConstraintChecker(
50, 4, np.inf, 4.0, 5, 5, 100, 8000, 8)
self.cnn_modifier = nn_modifiers.NNModifier(
self.cnn_constraint_checker)
self.mlp_modifier = nn_modifiers.NNModifier(
self.mlp_constraint_checker)
self.modifier_wo_cc = nn_modifiers.NNModifier(None)
def __init__(self, *args, **kwargs):
""" Constructor. """
super(DistSumNNKernelTestCase, self).__init__(*args, **kwargs)
self.non_assignment_penalty = 1
mlp_layer_labels, label_mismatch_penalty = \
nn_comparators.get_mlp_layer_label_mismatch_penalties(self.non_assignment_penalty,
'reg')
self.all_layer_labels = mlp_layer_labels
self.label_mismatch_penalty = label_mismatch_penalty
self.tp_comp = nn_comparators.OTMANNDistanceComputer(
mlp_layer_labels, label_mismatch_penalty,
self.non_assignment_penalty,
nn_comparators.MLP_STRUCTURAL_PENALTY_GROUPS,
nn_comparators.PATH_LENGTH_TYPES)
self.mislabel_coeffs = [2.0, 2.0, 1.0, 1.0, 1.0]
self.struct_coeffs = [0.25, 0.5, 1.0, 2.0, 4.0]
self.lp_betas = [1e-6] * len(self.struct_coeffs)
self.emd_betas = [1] * len(self.struct_coeffs)
self.mlps = generate_mlp_architectures()