def __init__(self, output_shape):
self._combiner = analyzers.WeightedMeanAndVarCombiner(
np.float32,
output_shape=output_shape,
compute_variance=False,
compute_weighted=True)
np.array([[1, 0, 0], [0, 1, 0], [0, 0, 1]], dtype=np.float64)
],
)
def _make_mean_and_var_accumulator_from_instance(instance, axis=None):
return analyzers._WeightedMeanAndVarAccumulator(
count=np.sum(np.ones_like(instance), axis=axis),
mean=np.mean(instance, axis=axis),
weight=np.sum(np.ones_like(instance), axis=axis),
variance=np.var(instance, axis=axis))
_MEAN_AND_VAR_TEST = dict(
testcase_name='WeightedMeanAndVar',
combiner=analyzers.WeightedMeanAndVarCombiner(np.float32, output_shape=()),
batches=[
_make_mean_and_var_accumulator_from_instance([[1, 2, 3, 4, 5, 6, 7]]),
# Count is 5*0xFFFF=327675 for this accumulator.
_make_mean_and_var_accumulator_from_instance([[8, 9, 10, 11, 12]] *
0xFFFF),
_make_mean_and_var_accumulator_from_instance([[100, 200, 3000]]),
],
expected_outputs=[
np.float32(10.00985092390558),
np.float32(29.418185761379473),
],
)
_MEAN_AND_VAR_BIG_TEST = dict(
testcase_name='WeightedMeanAndVarBig',
np.array([[1, 0, 0], [0, 1, 0], [0, 0, 1]], dtype=np.float64)
],
)
def _make_mean_and_var_accumulator_from_instance(instance, axis=None):
return analyzers._WeightedMeanAndVarAccumulator(
count=np.sum(np.ones_like(instance), axis=axis),
mean=np.mean(instance, axis=axis),
weight=np.sum(np.ones_like(instance), axis=axis),
variance=np.var(instance, axis=axis))
_MEAN_AND_VAR_TEST = dict(
testcase_name='WeightedMeanAndVar',
combiner=analyzers.WeightedMeanAndVarCombiner(np.float32, output_shape=()),
batches=[
_make_mean_and_var_accumulator_from_instance([[1, 2, 3, 4, 5, 6, 7]]),
# Count is 5*0xFFFF=327675 for this accumulator.
_make_mean_and_var_accumulator_from_instance([[8, 9, 10, 11, 12]] *
0xFFFF),
_make_mean_and_var_accumulator_from_instance([[100, 200, 3000]]),
],
expected_outputs=[
np.float32(10.00985092390558),
np.float32(29.418185761379473),
],
)
_MEAN_AND_VAR_SIMPLE_TEST = dict(
testcase_name='WeightedMeanAndVarSimple',
np.array([[1, 0, 0], [0, 1, 0], [0, 0, 1]], dtype=np.float64)
],
)
def _make_mean_and_var_accumulator_from_instance(instance, axis=None):
return analyzers._WeightedMeanAndVarAccumulator(
count=np.sum(np.ones_like(instance), axis=axis),
mean=np.mean(instance, axis=axis),
weight=1.,
variance=np.var(instance, axis=axis))
_MEAN_AND_VAR_TEST = dict(
testcase_name='WeightedMeanAndVar',
combiner=analyzers.WeightedMeanAndVarCombiner(np.float32),
batches=[
_make_mean_and_var_accumulator_from_instance([[1, 2, 3, 4, 5, 6, 7]]),
# Count is 5*0xFFFF=327675 for this accumulator.
_make_mean_and_var_accumulator_from_instance([[8, 9, 10, 11, 12]] *
0xFFFF),
_make_mean_and_var_accumulator_from_instance([[100, 200, 3000]]),
],
expected_outputs=[
np.float32(10.00985092390558),
np.float32(29.418185761379473),
],
)
_MEAN_AND_VAR_BIG_TEST = dict(
testcase_name='WeightedMeanAndVarBig',
