def test_check_target_array_2d_multiclass_good(self):
"""Ensures correct output from check_target_array.
In this case the input array is 2-D and multiclass, as expected.
"""
dl_utils.check_target_array(target_array=WIND_CLASS_MATRIX,
num_dimensions=2,
num_classes=WIND_CLASS_MATRIX.shape[1])
def test_check_target_array_2d_binary_good(self):
"""Ensures correct output from check_target_array.
In this case the input array is 2-D and contains 2 classes, as expected.
"""
dl_utils.check_target_array(target_array=TORNADO_CLASS_MATRIX,
num_dimensions=2,
num_classes=2)
def test_check_target_array_2d_bad_dim(self):
"""Ensures correct output from check_target_array.
In this case, the input array is 2-D but a 1-D array is expected.
"""
with self.assertRaises(TypeError):
dl_utils.check_target_array(target_array=TORNADO_CLASS_MATRIX,
num_dimensions=1,
num_classes=2)
def test_check_target_array_2d_multiclass_bad_class_num(self):
"""Ensures correct output from check_target_array.
In this case, the input array contains 6 classes but 2 classes are
expected.
"""
with self.assertRaises(TypeError):
dl_utils.check_target_array(target_array=WIND_CLASS_MATRIX,
num_dimensions=2,
num_classes=2)
def test_check_target_array_1d_binary_bad_class_num(self):
"""Ensures correct output from check_target_array.
In this case, 6 classes are expected and the input array contains only 2
classes. However, there is no way to ascertain that the 2-class array
is wrong (maybe higher classes just did not occur in the sample).
"""
dl_utils.check_target_array(target_array=TORNADO_CLASSES_1D,
num_dimensions=1,
num_classes=WIND_CLASS_MATRIX.shape[1])
def write_features(netcdf_file_name,
feature_matrix,
target_values,
num_classes,
append_to_file=False):
"""Writes features (activations of intermediate layer) to NetCDF file.
:param netcdf_file_name: Path to output file.
:param feature_matrix: numpy array of features. Must have >= 2 dimensions,
where the first dimension (length E) represents examples and the last
dimension represents channels (transformed input variables).
:param target_values: length-E numpy array of target values. Must all be
integers in 0...(K - 1), where K = number of classes.
:param num_classes: Number of classes.
:param append_to_file: Boolean flag. If True, will append to existing file.
If False, will create new file.
"""
error_checking.assert_is_boolean(append_to_file)
error_checking.assert_is_numpy_array(feature_matrix)
num_storm_objects = feature_matrix.shape[0]
dl_utils.check_target_array(target_array=target_values,
num_dimensions=1,
num_classes=num_classes)
error_checking.assert_is_numpy_array(target_values,
exact_dimensions=numpy.array(
[num_storm_objects]))
if append_to_file:
error_checking.assert_is_string(netcdf_file_name)
netcdf_dataset = netCDF4.Dataset(netcdf_file_name,
'a',
format='NETCDF3_64BIT_OFFSET')
prev_num_storm_objects = len(
numpy.array(netcdf_dataset.variables[TARGET_VALUES_KEY][:]))
netcdf_dataset.variables[FEATURE_MATRIX_KEY][prev_num_storm_objects:(
prev_num_storm_objects + num_storm_objects), ...] = feature_matrix
netcdf_dataset.variables[TARGET_VALUES_KEY][prev_num_storm_objects:(
prev_num_storm_objects + num_storm_objects)] = target_values
else:
file_system_utils.mkdir_recursive_if_necessary(
file_name=netcdf_file_name)
netcdf_dataset = netCDF4.Dataset(netcdf_file_name,
'w',
format='NETCDF3_64BIT_OFFSET')
netcdf_dataset.setncattr(NUM_CLASSES_KEY, num_classes)
netcdf_dataset.createDimension(STORM_OBJECT_DIMENSION_KEY, None)
netcdf_dataset.createDimension(FEATURE_DIMENSION_KEY,
feature_matrix.shape[1])
num_spatial_dimensions = len(feature_matrix.shape) - 2
tuple_of_dimension_keys = (STORM_OBJECT_DIMENSION_KEY, )
for i in range(num_spatial_dimensions):
netcdf_dataset.createDimension(SPATIAL_DIMENSION_KEYS[i],
feature_matrix.shape[i + 1])
tuple_of_dimension_keys += (SPATIAL_DIMENSION_KEYS[i], )
tuple_of_dimension_keys += (FEATURE_DIMENSION_KEY, )
netcdf_dataset.createVariable(FEATURE_MATRIX_KEY,
datatype=numpy.float32,
dimensions=tuple_of_dimension_keys)
netcdf_dataset.variables[FEATURE_MATRIX_KEY][:] = feature_matrix
netcdf_dataset.createVariable(TARGET_VALUES_KEY,
datatype=numpy.int32,
dimensions=STORM_OBJECT_DIMENSION_KEY)
netcdf_dataset.variables[TARGET_VALUES_KEY][:] = target_values
netcdf_dataset.close()
