def _get_unnormalized_pressure(model_metadata_dict, example_id_strings):
"""Returns profiles of unnormalized pressure.
E = number of examples
H = number of heights
:param model_metadata_dict: Dictionary read by `neural_net.read_metafile`.
:param example_id_strings: length-E list of example IDs.
:return: pressure_matrix_pascals: E-by-H numpy array of pressures.
"""
generator_option_dict = copy.deepcopy(
model_metadata_dict[neural_net.TRAINING_OPTIONS_KEY])
generator_option_dict[neural_net.SCALAR_PREDICTOR_NAMES_KEY] = [
example_utils.LATITUDE_NAME, example_utils.LONGITUDE_NAME,
example_utils.ZENITH_ANGLE_NAME
]
generator_option_dict[neural_net.VECTOR_PREDICTOR_NAMES_KEY] = [
example_utils.PRESSURE_NAME
]
generator_option_dict[neural_net.PREDICTOR_NORM_TYPE_KEY] = None
predictor_matrix = neural_net.create_data_specific_examples(
option_dict=generator_option_dict,
net_type_string=model_metadata_dict[neural_net.NET_TYPE_KEY],
example_id_strings=example_id_strings)[0]
dummy_example_dict = {
example_utils.SCALAR_PREDICTOR_NAMES_KEY:
generator_option_dict[neural_net.SCALAR_PREDICTOR_NAMES_KEY],
example_utils.VECTOR_PREDICTOR_NAMES_KEY:
generator_option_dict[neural_net.VECTOR_PREDICTOR_NAMES_KEY],
example_utils.HEIGHTS_KEY:
generator_option_dict[neural_net.HEIGHTS_KEY]
}
example_dict = neural_net.predictors_numpy_to_dict(
predictor_matrix=predictor_matrix,
example_dict=dummy_example_dict,
net_type_string=model_metadata_dict[neural_net.NET_TYPE_KEY])
return example_dict[example_utils.VECTOR_PREDICTOR_VALS_KEY][..., 0]
def test_predictors_numpy_to_dict_u_net(self):
"""Ensures correct output from predictors_numpy_to_dict.
In this case, neural-net type is U-net.
"""
this_example_dict = neural_net.predictors_numpy_to_dict(
predictor_matrix=U_NET_PREDICTOR_MATRIX,
example_dict=EXAMPLE_DICT,
net_type_string=neural_net.U_NET_TYPE_STRING)
self.assertTrue(
numpy.allclose(
this_example_dict[example_utils.VECTOR_PREDICTOR_VALS_KEY],
VECTOR_PREDICTOR_MATRIX,
atol=TOLERANCE))
self.assertTrue(
numpy.allclose(
this_example_dict[example_utils.SCALAR_PREDICTOR_VALS_KEY],
SCALAR_PREDICTOR_MATRIX,
atol=TOLERANCE))
def prediction_function(predictor_matrix):
"""Prediction function itself.
:param predictor_matrix: See doc for `run_forward_test_one_step`.
:return: prediction_matrices: 1-D list of numpy arrays, each containing
predicted values.
"""
net_type_string = model_metadata_dict[neural_net.NET_TYPE_KEY]
if net_type_string == neural_net.DENSE_NET_TYPE_STRING:
generator_option_dict = (
model_metadata_dict[neural_net.TRAINING_OPTIONS_KEY])
example_dict = {
example_utils.SCALAR_PREDICTOR_NAMES_KEY:
generator_option_dict[neural_net.SCALAR_PREDICTOR_NAMES_KEY],
example_utils.VECTOR_PREDICTOR_NAMES_KEY:
generator_option_dict[neural_net.VECTOR_PREDICTOR_NAMES_KEY],
example_utils.HEIGHTS_KEY:
generator_option_dict[neural_net.HEIGHTS_KEY]
}
new_example_dict = neural_net.predictors_numpy_to_dict(
predictor_matrix=predictor_matrix,
example_dict=example_dict,
net_type_string=neural_net.CNN_TYPE_STRING)
example_dict.update(new_example_dict)
this_predictor_matrix = neural_net.predictors_dict_to_numpy(
example_dict=example_dict, net_type_string=net_type_string)[0]
else:
this_predictor_matrix = predictor_matrix
return neural_net.apply_model(model_object=model_object,
predictor_matrix=this_predictor_matrix,
num_examples_per_batch=1000,
net_type_string=net_type_string,
is_loss_constrained_mse=False,
verbose=False)
def _run(model_file_name, example_file_name, num_examples, example_dir_name,
example_id_file_name, layer_name, is_layer_output, neuron_indices,
ideal_activation, output_file_name):
"""Makes saliency map for each example, according to one model.
This is effectively the main method.
:param model_file_name: See documentation at top of file.
:param example_file_name: Same.
:param num_examples: Same.
:param example_dir_name: Same.
:param example_id_file_name: Same.
:param layer_name: Same.
:param is_layer_output: Same.
:param neuron_indices: Same.
:param ideal_activation: Same.
:param output_file_name: Same.
"""
print('Reading model from: "{0:s}"...'.format(model_file_name))
model_object = neural_net.read_model(model_file_name)
metafile_name = neural_net.find_metafile(
model_dir_name=os.path.split(model_file_name)[0],
raise_error_if_missing=True
)
print('Reading metadata from: "{0:s}"...'.format(metafile_name))
metadata_dict = neural_net.read_metafile(metafile_name)
predictor_matrix, _, example_id_strings = (
misc_utils.get_examples_for_inference(
model_metadata_dict=metadata_dict,
example_file_name=example_file_name,
num_examples=num_examples, example_dir_name=example_dir_name,
example_id_file_name=example_id_file_name
)
)
print(SEPARATOR_STRING)
generator_option_dict = metadata_dict[neural_net.TRAINING_OPTIONS_KEY]
if is_layer_output:
dummy_example_dict = {
example_utils.SCALAR_TARGET_NAMES_KEY:
generator_option_dict[neural_net.SCALAR_TARGET_NAMES_KEY],
example_utils.VECTOR_TARGET_NAMES_KEY:
generator_option_dict[neural_net.VECTOR_TARGET_NAMES_KEY],
example_utils.HEIGHTS_KEY:
generator_option_dict[neural_net.HEIGHTS_KEY]
}
target_field_name, target_height_m_agl = (
neural_net.neuron_indices_to_target_var(
neuron_indices=neuron_indices,
example_dict=copy.deepcopy(dummy_example_dict),
net_type_string=metadata_dict[neural_net.NET_TYPE_KEY]
)
)
else:
target_field_name = None
target_height_m_agl = None
print('Target field and height = {0:s}, {1:s}'.format(
str(target_field_name), str(target_height_m_agl)
))
print('Computing saliency for neuron {0:s} in layer "{1:s}"...'.format(
str(neuron_indices), layer_name
))
saliency_matrix = saliency.get_saliency_one_neuron(
model_object=model_object, predictor_matrix=predictor_matrix,
layer_name=layer_name, neuron_indices=neuron_indices,
ideal_activation=ideal_activation
)
net_type_string = metadata_dict[neural_net.NET_TYPE_KEY]
if net_type_string == neural_net.DENSE_NET_TYPE_STRING:
dummy_example_dict = {
example_utils.SCALAR_PREDICTOR_NAMES_KEY:
generator_option_dict[neural_net.SCALAR_PREDICTOR_NAMES_KEY],
example_utils.VECTOR_PREDICTOR_NAMES_KEY:
generator_option_dict[neural_net.VECTOR_PREDICTOR_NAMES_KEY],
example_utils.HEIGHTS_KEY:
generator_option_dict[neural_net.HEIGHTS_KEY]
}
dummy_example_dict = neural_net.predictors_numpy_to_dict(
predictor_matrix=saliency_matrix, example_dict=dummy_example_dict,
net_type_string=metadata_dict[neural_net.NET_TYPE_KEY]
)
scalar_saliency_matrix = (
dummy_example_dict[example_utils.SCALAR_PREDICTOR_VALS_KEY]
)
vector_saliency_matrix = (
dummy_example_dict[example_utils.VECTOR_PREDICTOR_VALS_KEY]
)
else:
num_scalar_predictors = len(
generator_option_dict[neural_net.SCALAR_PREDICTOR_NAMES_KEY]
)
scalar_saliency_matrix = saliency_matrix[..., -num_scalar_predictors:]
vector_saliency_matrix = saliency_matrix[..., :-num_scalar_predictors]
print('Writing saliency maps to: "{0:s}"...'.format(output_file_name))
saliency.write_file(
netcdf_file_name=output_file_name,
scalar_saliency_matrix=scalar_saliency_matrix,
vector_saliency_matrix=vector_saliency_matrix,
example_id_strings=example_id_strings, model_file_name=model_file_name,
layer_name=layer_name, neuron_indices=neuron_indices,
ideal_activation=ideal_activation, target_field_name=target_field_name,
target_height_m_agl=target_height_m_agl
)
def _run(model_file_name, example_file_name, num_examples, example_dir_name,
example_id_file_name, ideal_activation, scalar_output_layer_name,
vector_output_layer_name, output_file_name):
"""Makes saliency map for each example and target, according to one model.
This is effectively the main method.
:param model_file_name: See documentation at top of file.
:param example_file_name: Same.
:param num_examples: Same.
:param example_dir_name: Same.
:param example_id_file_name: Same.
:param ideal_activation: Same.
:param scalar_output_layer_name: Same.
:param vector_output_layer_name: Same.
:param output_file_name: Same.
"""
print('Reading model from: "{0:s}"...'.format(model_file_name))
model_object = neural_net.read_model(model_file_name)
metafile_name = neural_net.find_metafile(
model_dir_name=os.path.split(model_file_name)[0],
raise_error_if_missing=True
)
print('Reading metadata from: "{0:s}"...'.format(metafile_name))
metadata_dict = neural_net.read_metafile(metafile_name)
predictor_matrix, _, example_id_strings = (
misc_utils.get_examples_for_inference(
model_metadata_dict=metadata_dict,
example_file_name=example_file_name,
num_examples=num_examples, example_dir_name=example_dir_name,
example_id_file_name=example_id_file_name
)
)
print(SEPARATOR_STRING)
net_type_string = metadata_dict[neural_net.NET_TYPE_KEY]
generator_option_dict = metadata_dict[neural_net.TRAINING_OPTIONS_KEY]
scalar_predictor_names = (
generator_option_dict[neural_net.SCALAR_PREDICTOR_NAMES_KEY]
)
vector_predictor_names = (
generator_option_dict[neural_net.VECTOR_PREDICTOR_NAMES_KEY]
)
scalar_target_names = (
generator_option_dict[neural_net.SCALAR_TARGET_NAMES_KEY]
)
vector_target_names = (
generator_option_dict[neural_net.VECTOR_TARGET_NAMES_KEY]
)
heights_m_agl = generator_option_dict[neural_net.HEIGHTS_KEY]
num_scalar_predictors = len(scalar_predictor_names)
num_vector_predictors = len(vector_predictor_names)
num_scalar_targets = len(scalar_target_names)
num_vector_targets = len(vector_target_names)
num_heights = len(heights_m_agl)
num_examples = len(example_id_strings)
if net_type_string == neural_net.DENSE_NET_TYPE_STRING:
saliency_matrix_scalar_p_scalar_t = numpy.full(
(num_examples, num_scalar_predictors, num_scalar_targets), numpy.nan
)
else:
saliency_matrix_scalar_p_scalar_t = numpy.full(
(num_examples, num_heights, num_scalar_predictors,
num_scalar_targets),
numpy.nan
)
saliency_matrix_vector_p_scalar_t = numpy.full(
(num_examples, num_heights, num_vector_predictors, num_scalar_targets),
numpy.nan
)
if net_type_string == neural_net.DENSE_NET_TYPE_STRING:
saliency_matrix_scalar_p_vector_t = numpy.full(
(num_examples, num_scalar_predictors, num_heights,
num_vector_targets),
numpy.nan
)
else:
saliency_matrix_scalar_p_vector_t = numpy.full(
(num_examples, num_heights, num_scalar_predictors, num_heights,
num_vector_targets),
numpy.nan
)
saliency_matrix_vector_p_vector_t = numpy.full(
(num_examples, num_heights, num_vector_predictors, num_heights,
num_vector_targets),
numpy.nan
)
dummy_example_dict = {
example_utils.SCALAR_PREDICTOR_NAMES_KEY: scalar_predictor_names,
example_utils.VECTOR_PREDICTOR_NAMES_KEY: vector_predictor_names,
example_utils.SCALAR_TARGET_NAMES_KEY: scalar_target_names,
example_utils.VECTOR_TARGET_NAMES_KEY: vector_target_names,
example_utils.HEIGHTS_KEY: heights_m_agl
}
for k in range(num_scalar_targets):
these_neuron_indices = neural_net.target_var_to_neuron_indices(
example_dict=copy.deepcopy(dummy_example_dict),
net_type_string=net_type_string, target_name=scalar_target_names[k]
)
print('Computing saliency for "{0:s}"...'.format(
scalar_target_names[k]
))
this_saliency_matrix = saliency.get_saliency_one_neuron(
model_object=model_object, predictor_matrix=predictor_matrix,
layer_name=scalar_output_layer_name,
neuron_indices=these_neuron_indices,
ideal_activation=ideal_activation
)
if net_type_string == neural_net.DENSE_NET_TYPE_STRING:
new_example_dict = neural_net.predictors_numpy_to_dict(
predictor_matrix=this_saliency_matrix,
example_dict=copy.deepcopy(dummy_example_dict),
net_type_string=net_type_string
)
saliency_matrix_scalar_p_scalar_t[..., k] = (
new_example_dict[example_utils.SCALAR_PREDICTOR_VALS_KEY]
)
saliency_matrix_vector_p_scalar_t[..., k] = (
new_example_dict[example_utils.VECTOR_PREDICTOR_VALS_KEY]
)
else:
saliency_matrix_scalar_p_scalar_t[..., k] = (
this_saliency_matrix[..., -num_scalar_predictors:]
)
saliency_matrix_vector_p_scalar_t[..., k] = (
this_saliency_matrix[..., :-num_scalar_predictors]
)
print(SEPARATOR_STRING)
for k in range(num_vector_targets):
for j in range(num_heights):
these_neuron_indices = neural_net.target_var_to_neuron_indices(
example_dict=copy.deepcopy(dummy_example_dict),
net_type_string=net_type_string,
target_name=vector_target_names[k],
height_m_agl=heights_m_agl[j]
)
print('Computing saliency for "{0:s}" at {1:d} m AGL...'.format(
vector_target_names[k], int(numpy.round(heights_m_agl[j]))
))
this_saliency_matrix = saliency.get_saliency_one_neuron(
model_object=model_object, predictor_matrix=predictor_matrix,
layer_name=vector_output_layer_name,
neuron_indices=these_neuron_indices,
ideal_activation=ideal_activation
)
if net_type_string == neural_net.DENSE_NET_TYPE_STRING:
new_example_dict = neural_net.predictors_numpy_to_dict(
predictor_matrix=this_saliency_matrix,
example_dict=copy.deepcopy(dummy_example_dict),
net_type_string=net_type_string
)
saliency_matrix_scalar_p_vector_t[..., j, k] = (
new_example_dict[example_utils.SCALAR_PREDICTOR_VALS_KEY]
)
saliency_matrix_vector_p_vector_t[..., j, k] = (
new_example_dict[example_utils.VECTOR_PREDICTOR_VALS_KEY]
)
else:
saliency_matrix_scalar_p_vector_t[..., j, k] = (
this_saliency_matrix[..., -num_scalar_predictors:]
)
saliency_matrix_vector_p_vector_t[..., j, k] = (
this_saliency_matrix[..., :-num_scalar_predictors]
)
print(SEPARATOR_STRING)
print('Writing saliency maps to: "{0:s}"...'.format(output_file_name))
saliency.write_all_targets_file(
netcdf_file_name=output_file_name,
saliency_matrix_scalar_p_scalar_t=saliency_matrix_scalar_p_scalar_t,
saliency_matrix_vector_p_scalar_t=saliency_matrix_vector_p_scalar_t,
saliency_matrix_scalar_p_vector_t=saliency_matrix_scalar_p_vector_t,
saliency_matrix_vector_p_vector_t=saliency_matrix_vector_p_vector_t,
example_id_strings=example_id_strings, model_file_name=model_file_name,
ideal_activation=ideal_activation
)
def _run(model_file_name, example_file_name, num_examples, example_dir_name,
example_id_file_name, layer_name, neuron_indices, ideal_activation,
num_iterations, learning_rate, l2_weight, output_file_name):
"""Runs backwards optimization.
This is effectively the main method.
:param model_file_name: See documentation at top of file.
:param example_file_name: Same.
:param num_examples: Same.
:param example_dir_name: Same.
:param example_id_file_name: Same.
:param layer_name: Same.
:param neuron_indices: Same.
:param ideal_activation: Same.
:param num_iterations: Same.
:param learning_rate: Same.
:param l2_weight: Same.
:param output_file_name: Same.
"""
print('Reading model from: "{0:s}"...'.format(model_file_name))
model_object = neural_net.read_model(model_file_name)
metafile_name = neural_net.find_metafile(
model_dir_name=os.path.split(model_file_name)[0],
raise_error_if_missing=True
)
print('Reading metadata from: "{0:s}"...'.format(metafile_name))
metadata_dict = neural_net.read_metafile(metafile_name)
predictor_matrix, _, example_id_strings = (
misc_utils.get_examples_for_inference(
model_metadata_dict=metadata_dict,
example_file_name=example_file_name,
num_examples=num_examples, example_dir_name=example_dir_name,
example_id_file_name=example_id_file_name
)
)
print(SEPARATOR_STRING)
generator_option_dict = metadata_dict[neural_net.TRAINING_OPTIONS_KEY]
normalization_file_name = (
generator_option_dict[neural_net.NORMALIZATION_FILE_KEY]
)
print((
'Reading training examples (for normalization) from: "{0:s}"...'
).format(
normalization_file_name
))
training_example_dict = example_io.read_file(normalization_file_name)
training_example_dict = example_utils.subset_by_height(
example_dict=training_example_dict,
heights_m_agl=generator_option_dict[neural_net.HEIGHTS_KEY]
)
num_examples = len(example_id_strings)
bwo_dict = None
for i in range(num_examples):
this_bwo_dict = bwo.optimize_input_for_neuron(
model_object=model_object,
init_function_or_matrix=predictor_matrix[i, ...],
layer_name=layer_name, neuron_indices=neuron_indices,
ideal_activation=ideal_activation, num_iterations=num_iterations,
learning_rate=learning_rate, l2_weight=l2_weight
)
if i == num_examples - 1:
print(SEPARATOR_STRING)
else:
print(MINOR_SEPARATOR_STRING)
if bwo_dict is None:
these_dim = numpy.array(
(num_examples,) +
this_bwo_dict[bwo.INITIAL_PREDICTORS_KEY].shape[1:],
dtype=int
)
bwo_dict = {
bwo.INITIAL_PREDICTORS_KEY: numpy.full(these_dim, numpy.nan),
bwo.FINAL_PREDICTORS_KEY: numpy.full(these_dim, numpy.nan),
bwo.INITIAL_ACTIVATIONS_KEY:
numpy.full(num_examples, numpy.nan),
bwo.FINAL_ACTIVATIONS_KEY: numpy.full(num_examples, numpy.nan)
}
bwo_dict[bwo.INITIAL_PREDICTORS_KEY][i, ...] = (
this_bwo_dict[bwo.INITIAL_PREDICTORS_KEY][0, ...]
)
bwo_dict[bwo.FINAL_PREDICTORS_KEY][i, ...] = (
this_bwo_dict[bwo.FINAL_PREDICTORS_KEY][0, ...]
)
bwo_dict[bwo.INITIAL_ACTIVATIONS_KEY][i] = (
this_bwo_dict[bwo.INITIAL_ACTIVATION_KEY]
)
bwo_dict[bwo.FINAL_ACTIVATIONS_KEY][i] = (
this_bwo_dict[bwo.FINAL_ACTIVATION_KEY]
)
if example_file_name == '':
example_file_name = example_io.find_many_files(
directory_name=example_dir_name,
first_time_unix_sec=0, last_time_unix_sec=int(1e12),
raise_error_if_any_missing=False, raise_error_if_all_missing=True
)[0]
first_example_dict = example_io.read_file(example_file_name)
first_example_dict = example_utils.subset_by_height(
example_dict=first_example_dict,
heights_m_agl=generator_option_dict[neural_net.HEIGHTS_KEY]
)
net_type_string = metadata_dict[neural_net.NET_TYPE_KEY]
init_example_dict = copy.deepcopy(first_example_dict)
this_example_dict = neural_net.predictors_numpy_to_dict(
predictor_matrix=bwo_dict[bwo.INITIAL_PREDICTORS_KEY],
example_dict=init_example_dict, net_type_string=net_type_string
)
init_example_dict.update(this_example_dict)
if generator_option_dict[neural_net.PREDICTOR_NORM_TYPE_KEY] is not None:
init_example_dict = normalization.denormalize_data(
new_example_dict=init_example_dict,
training_example_dict=training_example_dict,
normalization_type_string=
generator_option_dict[neural_net.PREDICTOR_NORM_TYPE_KEY],
min_normalized_value=
generator_option_dict[neural_net.PREDICTOR_MIN_NORM_VALUE_KEY],
max_normalized_value=
generator_option_dict[neural_net.PREDICTOR_MAX_NORM_VALUE_KEY],
separate_heights=True, apply_to_predictors=True,
apply_to_vector_targets=False, apply_to_scalar_targets=False
)
init_scalar_predictor_matrix = (
init_example_dict[example_utils.SCALAR_PREDICTOR_VALS_KEY]
)
init_vector_predictor_matrix = (
init_example_dict[example_utils.VECTOR_PREDICTOR_VALS_KEY]
)
final_example_dict = copy.deepcopy(first_example_dict)
this_example_dict = neural_net.predictors_numpy_to_dict(
predictor_matrix=bwo_dict[bwo.FINAL_PREDICTORS_KEY],
example_dict=final_example_dict, net_type_string=net_type_string
)
final_example_dict.update(this_example_dict)
if generator_option_dict[neural_net.PREDICTOR_NORM_TYPE_KEY] is not None:
final_example_dict = normalization.denormalize_data(
new_example_dict=final_example_dict,
training_example_dict=training_example_dict,
normalization_type_string=
generator_option_dict[neural_net.PREDICTOR_NORM_TYPE_KEY],
min_normalized_value=
generator_option_dict[neural_net.PREDICTOR_MIN_NORM_VALUE_KEY],
max_normalized_value=
generator_option_dict[neural_net.PREDICTOR_MAX_NORM_VALUE_KEY],
separate_heights=True, apply_to_predictors=True,
apply_to_vector_targets=False, apply_to_scalar_targets=False
)
final_scalar_predictor_matrix = (
final_example_dict[example_utils.SCALAR_PREDICTOR_VALS_KEY]
)
final_vector_predictor_matrix = (
final_example_dict[example_utils.VECTOR_PREDICTOR_VALS_KEY]
)
print('Writing results to file: "{0:s}"...'.format(output_file_name))
bwo.write_file(
netcdf_file_name=output_file_name,
init_scalar_predictor_matrix=init_scalar_predictor_matrix,
final_scalar_predictor_matrix=final_scalar_predictor_matrix,
init_vector_predictor_matrix=init_vector_predictor_matrix,
final_vector_predictor_matrix=final_vector_predictor_matrix,
initial_activations=bwo_dict[bwo.INITIAL_ACTIVATIONS_KEY],
final_activations=bwo_dict[bwo.FINAL_ACTIVATIONS_KEY],
example_id_strings=example_id_strings, model_file_name=model_file_name,
layer_name=layer_name, neuron_indices=neuron_indices,
ideal_activation=ideal_activation, num_iterations=num_iterations,
learning_rate=learning_rate, l2_weight=l2_weight
)
def run_backwards_test(predictor_matrix,
target_matrices,
model_object,
model_metadata_dict,
cost_function,
shuffle_profiles_together=True,
num_bootstrap_reps=DEFAULT_NUM_BOOTSTRAP_REPS):
"""Runs backwards version of permutation test (both single- and multi-pass).
B = number of bootstrap replicates
N = number of predictors (either channels or height/channel pairs) available
to permute
:param predictor_matrix: See doc for `run_forward_test`.
:param target_matrices: Same.
:param model_object: Same.
:param model_metadata_dict: Same.
:param cost_function: Same.
:param shuffle_profiles_together: Same.
:param num_bootstrap_reps: Same.
:return: result_dict: Dictionary with the following keys.
result_dict['orig_cost_estimates']: length-B numpy array with estimates of
original cost (before *de*permutation).
result_dict['best_predictor_names']: length-N list with best predictor at
each step.
result_dict['best_heights_m_agl']: length-N numpy array of corresponding
heights (metres above ground level). This may be None.
result_dict['best_cost_matrix']: N-by-B numpy array of costs after
*de*permutation at each step.
result_dict['step1_predictor_names']: length-N list with predictors in order
that they were *de*permuted in step 1.
result_dict['step1_heights_m_agl']: length-N numpy array of corresponding
heights (metres above ground level). This may be None.
result_dict['step1_cost_matrix']: N-by-B numpy array of costs after
*de*permutation in step 1.
result_dict['is_backwards_test']: Boolean flag (always True for this
method).
"""
error_checking.assert_is_boolean(shuffle_profiles_together)
error_checking.assert_is_integer(num_bootstrap_reps)
num_bootstrap_reps = numpy.maximum(num_bootstrap_reps, 1)
generator_option_dict = model_metadata_dict[
neural_net.TRAINING_OPTIONS_KEY]
example_dict = {
example_utils.SCALAR_PREDICTOR_NAMES_KEY:
generator_option_dict[neural_net.SCALAR_PREDICTOR_NAMES_KEY],
example_utils.VECTOR_PREDICTOR_NAMES_KEY:
generator_option_dict[neural_net.VECTOR_PREDICTOR_NAMES_KEY],
example_utils.HEIGHTS_KEY:
generator_option_dict[neural_net.HEIGHTS_KEY]
}
new_example_dict = neural_net.predictors_numpy_to_dict(
predictor_matrix=predictor_matrix,
example_dict=example_dict,
net_type_string=model_metadata_dict[neural_net.NET_TYPE_KEY])
example_dict.update(new_example_dict)
predictor_matrix, predictor_name_matrix, height_matrix_m_agl = (
neural_net.predictors_dict_to_numpy(
example_dict=example_dict,
net_type_string=neural_net.CNN_TYPE_STRING))
scalar_channel_flags = numpy.isnan(height_matrix_m_agl[0, :])
prediction_function = _make_prediction_function(
model_object=model_object, model_metadata_dict=model_metadata_dict)
# Permute all predictors.
num_channels = predictor_matrix.shape[-1]
if shuffle_profiles_together:
num_heights = 1
else:
num_heights = predictor_matrix.shape[-2]
clean_predictor_matrix = predictor_matrix + 0.
for j in range(num_heights):
for k in range(num_channels):
predictor_matrix = _permute_values(
predictor_matrix=predictor_matrix,
channel_index=k,
height_index=None if shuffle_profiles_together else j)[0]
# Find original cost (before *de*permutation).
print('Finding original cost (before *de*permutation)...')
orig_cost_estimates = _bootstrap_cost(
target_matrices=target_matrices,
prediction_matrices=prediction_function(predictor_matrix),
cost_function=cost_function,
num_replicates=num_bootstrap_reps)
# Do dirty work.
permuted_flag_matrix = numpy.full(predictor_matrix.shape[1:],
True,
dtype=bool)
best_predictor_names = []
best_cost_matrix = numpy.full((0, num_bootstrap_reps), numpy.nan)
if shuffle_profiles_together:
best_heights_m_agl = None
else:
best_heights_m_agl = []
step1_predictor_names = None
step1_heights_m_agl = None
step1_cost_matrix = None
step_num = 0
while True:
print(MINOR_SEPARATOR_STRING)
step_num += 1
this_result_dict = run_backwards_test_one_step(
predictor_matrix=predictor_matrix,
clean_predictor_matrix=clean_predictor_matrix,
target_matrices=target_matrices,
prediction_function=prediction_function,
cost_function=cost_function,
permuted_flag_matrix=permuted_flag_matrix,
scalar_channel_flags=scalar_channel_flags,
shuffle_profiles_together=shuffle_profiles_together,
num_bootstrap_reps=num_bootstrap_reps)
if this_result_dict is None:
break
predictor_matrix = this_result_dict[PREDICTORS_KEY]
permuted_flag_matrix = this_result_dict[PERMUTED_FLAGS_KEY]
these_predictor_names, these_heights_m_agl = (
_predictor_indices_to_metadata(
all_predictor_name_matrix=predictor_name_matrix,
all_height_matrix_m_agl=height_matrix_m_agl,
one_step_result_dict=this_result_dict))
this_best_index = numpy.argmin(
numpy.mean(this_result_dict[DEPERMUTED_COSTS_KEY], axis=1))
best_predictor_names.append(these_predictor_names[this_best_index])
best_cost_matrix = numpy.concatenate(
(best_cost_matrix,
this_result_dict[DEPERMUTED_COSTS_KEY][[this_best_index], :]),
axis=0)
log_string = 'Best predictor at {0:d}th step = {1:s}'.format(
step_num, best_predictor_names[-1])
if not shuffle_profiles_together:
best_heights_m_agl.append(these_heights_m_agl[this_best_index])
log_string += ' at {0:s} m AGL'.format(best_heights_m_agl[-1])
log_string += ' ... cost = {0:f}'.format(
numpy.mean(best_cost_matrix[-1, :]))
print(log_string)
if step_num != 1:
continue
step1_predictor_names = copy.deepcopy(these_predictor_names)
step1_cost_matrix = this_result_dict[DEPERMUTED_COSTS_KEY] + 0.
if not shuffle_profiles_together:
step1_heights_m_agl = these_heights_m_agl + 0
return {
ORIGINAL_COST_KEY: orig_cost_estimates,
BEST_PREDICTORS_KEY: best_predictor_names,
BEST_HEIGHTS_KEY: best_heights_m_agl,
BEST_COSTS_KEY: best_cost_matrix,
STEP1_PREDICTORS_KEY: step1_predictor_names,
STEP1_HEIGHTS_KEY: step1_heights_m_agl,
STEP1_COSTS_KEY: step1_cost_matrix,
BACKWARDS_FLAG_KEY: True
}
