def test_shape_transformer_2d_2d_array(axis):
st = ShapeTransformerTo2D(axis=axis)
x = numpy.arange(6).reshape((2, 3))
xt = st.transform(x)
xit = st.inverse_transform(xt)
if axis == 0:
assert numpy.allclose(xt.array, numpy.array([[0, 3], [1, 4], [2, 5]]))
elif axis == 1 or axis == -1:
assert numpy.allclose(x, xt.array)
assert numpy.allclose(x, xit.array)
def transform(self, x, axis=1):
is_array = not isinstance(x, Variable)
if self.mean is None:
raise AttributeError('[Error] mean is None, call fit beforehand!')
x = format_x(x)
shape_transformer = ShapeTransformerTo2D(axis=axis)
x = shape_transformer.transform(x)
mean_all, std_all = self._compute_mean_std_all(x.shape[1])
x = (x - mean_all[None, :]) / std_all[None, :]
x = shape_transformer.inverse_transform(x)
if is_array:
x = x.array
return x
def test_shape_transformer_2d_3d_array(axis):
st = ShapeTransformerTo2D(axis=axis)
x = numpy.arange(12).reshape((2, 3, 2))
xt = st.transform(x)
xit = st.inverse_transform(xt)
if axis == 0:
assert numpy.allclose(
xt.array,
numpy.array([[0, 6], [1, 7], [2, 8], [3, 9], [4, 10], [5, 11]]))
elif axis == 1:
assert numpy.allclose(
xt.array,
numpy.array([[0, 2, 4], [1, 3, 5], [6, 8, 10], [7, 9, 11]]))
elif axis == 2 or axis == -1:
assert numpy.allclose(xt.array, x.reshape(6, 2))
assert numpy.allclose(x, xit.array)
def fit(self, x, indices=None, axis=1):
"""Fitting parameter.
Args:
x (numpy.ndarray or cupy.ndarray or Variable):
indices (list or tuple or None):
indices for applying standard scaling.
axis (int): axis to calculate mean & std.
Returns:
self (StandardScaler): this instance.
"""
x = to_array(x)
x = format_x(x)
x = ShapeTransformerTo2D(axis=axis).transform(x).array
if indices is None:
pass
elif isinstance(indices, (list, tuple)):
indices = numpy.asarray(indices)
self.indices = indices
if self.indices is not None:
x = x[:, self.indices]
xp = self.xp
if xp is numpy:
x = cuda.to_cpu(x)
self.mean = xp.nanmean(x, axis=0)
self.std = xp.nanstd(x, axis=0)
else:
x = cuda.to_gpu(x)
if int(xp.sum(xp.isnan(x))) > 0:
raise NotImplementedError(
"StandardScaling with nan value on GPU is not supported.")
# cupy.nanmean, cupy.nanstd is not implemented yet.
self.mean = xp.mean(x, axis=0)
self.std = xp.std(x, axis=0)
# result consistency check
if xp.sum(self.std == 0) > 0:
logger = getLogger(__name__)
ind = numpy.argwhere(cuda.to_cpu(self.std) == 0)[:, 0]
logger.warning('fit: std was 0 at indices {}'.format(ind))
return self
def fit(self, x, indices=None, axis=1):
"""Fitting parameter.
Args:
x (numpy.ndarray or cupy.ndarray or Variable):
indices (list or tuple or None):
indices for applying standard scaling.
axis (int): axis to calculate min & max.
Returns:
self (MinMaxScaler): this instance.
"""
x = to_array(x)
x = format_x(x)
x = ShapeTransformerTo2D(axis=axis).transform(x).array
if indices is None:
pass
elif isinstance(indices, (list, tuple)):
indices = numpy.asarray(indices)
self.indices = indices
if self.indices is not None:
x = x[:, self.indices]
xp = self.xp
if xp is numpy:
x = cuda.to_cpu(x)
else:
x = cuda.to_gpu(x)
self.min = xp.nanmin(x, axis=0)
self.max = xp.nanmax(x, axis=0)
# result consistency check
if xp.sum(self.max - self.min == 0) > 0:
logger = getLogger(__name__)
ind = numpy.argwhere(cuda.to_cpu(self.max-self.min) == 0)[:, 0]
logger.warning('fit: max-min was 0 at indices {}'.format(ind))
return self
def test_shape_transformer_2d_error():
st = ShapeTransformerTo2D(axis=1)
x = numpy.arange(6).reshape(2, 3)
with pytest.raises(AttributeError):
# call `inverse_transform` before `transform`
xt = st.inverse_transform(x)