def generate_data_set():
while True:
b0 = random.randint(0, 140)
b1 = random.randint(0, 140)
d0 = min(b0 + random.randint(0, 127), 255)
d1 = min(b1 + random.randint(0, 127), 255)
label = random.choice([0, 1])
if label == 0:
yield np.array([b0, d0, b1, d1]), 0
else:
yield np.array([d0, b0, d1, b1]), 1
def test_setitem_and_copy(tt):
ttc = copy(tt)
if len(ttc) > 0:
a = ttc[0]
b = ttc[-1]
ttc[0] = np.array([1, 2, 3])
ttc[-1] = np.array([1, 2, 3])
ttc[0] = a
ttc[-1] = b
assert ttc == tt
with pytest.raises(IndexError):
ttc[len(ttc)]
def save_data_set(n: int, file_path):
trainX, trainY = zip(*list(islice(generate_data_set(), n)))
testX, testY = zip(*list(islice(generate_data_set(), n)))
trainX = np.array(list(trainX))
trainY = np.array(list(trainY))
testX = np.array(list(testX))
testY = np.array(list(testY))
savemat(file_path, {
'trainX': trainX,
'trainY': trainY,
'testX': testX,
'testY': testY,
},
do_compression=True)
def shrink(images):
return np.array([
shrink_quadratic(img, new_side_length).flatten() for img in images
])
from tensor_networks.transposition import transpose_outer_indices, reverse_dimensions
from tests.helpers import constant_fixture, arange_from_shape
from tensor_networks.patched_numpy import np
arr = constant_fixture(params=[
arange_from_shape(9, 9),
np.random.random((3, 7, 5, 3, 2)),
np.array([]),
arange_from_shape(8),
])
def test_transpose_bond_indices(arr):
assert (arr == transpose_outer_indices(transpose_outer_indices(arr))).all()
def test_reverse_transpose(arr):
assert (arr == reverse_dimensions(reverse_dimensions(arr))).all()
def image_feature_linear(absolute_colors: Array) -> Array:
return np.array(
list(
map(greyscale_feature_linear,
map(color_abs_to_percentage, absolute_colors))))
def greyscale_feature_linear(percentage: PartialColor) -> Array:
"""
:param percentage: A grey value
:return: An array of a black value and a white value with a sum of 1
"""
return np.array([percentage, 1 - percentage])
def greyscale_feature_sin_cos(percentage: PartialColor) -> Array:
"""
:param percentage: A grey value
:return: An array of a black value and a white value with a sum of 1
"""
return np.array([cos(pi / 2 * percentage), sin(pi / 2 * percentage)])
def test_arithmetic(tt):
assert np.array(tt.cores) - np.array(((tt / 5) * 5).cores) == approx(0)
assert np.array(tt.cores) - np.array(((5 * tt) / 5).cores) == approx(0)