def test__convert_dataset(self):
# Initialize test values
d = np.arange(10)
d = np.sin(d) + 1j * np.cos(d)
d_re = np.reshape(np.real(d), d.shape + (1,))
# Test multiple data sets
r = Complex2Features._convert_dataset((d, d), True, False, False, False)
npt.assert_array_almost_equal(r, np.concatenate((d_re, d_re), axis=-1))
def test__convert_data_entry(self):
# Test valid result shape
d = np.ones((10,256))
r = Complex2Features._convert_data_entry(d, True, False, True, False)
self.assertEqual(r.shape, (10,256,2))
d = np.ones((17,128,9))
r = Complex2Features._convert_data_entry(d, False, True, True, True)
self.assertEqual(r.shape, (17,128,9,3))
# Initialize test values
d = np.arange(10)
d = np.sin(d) + 1j * np.cos(d)
d_re = np.reshape(np.real(d), d.shape + (1,))
d_im = np.reshape(np.imag(d), d.shape + (1,))
d_abs = np.reshape(np.abs(d), d.shape + (1,))
d_phi = np.reshape(np.angle(d), d.shape + (1,))
# Test single valued results
r = Complex2Features._convert_data_entry(d, True, False, False, False)
npt.assert_array_almost_equal(r, d_re)
r = Complex2Features._convert_data_entry(d, False, True, False, False)
npt.assert_array_almost_equal(r, d_im)
r = Complex2Features._convert_data_entry(d, False, False, True, False)
npt.assert_array_almost_equal(r, d_abs)
r = Complex2Features._convert_data_entry(d, False, False, False, True)
npt.assert_array_almost_equal(r, d_phi)
# Test multiple valued result
r = Complex2Features._convert_data_entry(d, True, True, True, False)
npt.assert_array_almost_equal(r, np.concatenate((d_re, d_im, d_abs), axis=-1))
def test__assemble_dataset(self):
d1 = np.ones((10, 20))
d2 = np.zeros((10, 20))
d3 = np.reshape(np.arange(200), (10,20))
r = Complex2Features._assemble_dataset((d1, d2, d3))
self.assertTupleEqual(r.shape, (10,20,3))
self.assertEqual(r[5, 17, 0], 1)
self.assertEqual(r[7, 3, 1], 0)
self.assertEqual(r[0, 13, 2], 13)
def test_apply(self):
for grouped in (True, False):
# Test for valid shapes
d = np.ones((10,256))
r = Complex2Features.apply(d, False, True, True, False, False, grouped)
self.assertEqual(r.shape, (10,256,2))
r = Complex2Features.apply((d, -d), True, True, True, False, True, grouped)
self.assertEqual(r.shape, (10,256,6))
r = Complex2Features.apply((d, d, -2*d), 2, True, False, False, False, grouped)
self.assertEqual(r.shape, (10,256,3))
# Test values
d = np.reshape(np.arange(2560), (10,256))
d = np.sin(d) + 1j * np.cos(d)
d_re = np.reshape(np.real(d), d.shape)
r = Complex2Features.apply((d, 2*d), False, True, False, False, False, grouped)
self.assertEqual(r.shape, (10,256,2))
npt.assert_array_almost_equal(d_re, r[...,0])
npt.assert_array_almost_equal(2*d_re, r[...,1])
# Test ordering
d = np.reshape(np.arange(2560), (10,256))
d = np.sin(d) + 1j * np.cos(d)
d_re = np.reshape(np.real(d), d.shape)
d_im = np.reshape(np.imag(d), d.shape)
r = Complex2Features.apply((d, 2*d), False, True, True, False, False, False)
self.assertEqual(r.shape, (10, 256, 4))
npt.assert_array_almost_equal(d_re, r[..., 0])
npt.assert_array_almost_equal(2 * d_re, r[..., 1])
npt.assert_array_almost_equal(d_im, r[..., 2])
npt.assert_array_almost_equal(2 * d_im, r[..., 3])
r = Complex2Features.apply((d, 2*d), False, True, True, False, False, True)
self.assertEqual(r.shape, (10, 256, 4))
npt.assert_array_almost_equal(d_re, r[..., 0])
npt.assert_array_almost_equal(2 * d_re, r[..., 2])
npt.assert_array_almost_equal(d_im, r[..., 1])
npt.assert_array_almost_equal(2 * d_im, r[..., 3])
def test__normalize_dataset(self):
d_ones = np.ones((10,256))
d_arange = np.reshape(np.arange(2560).astype(float), (10,256))
d_max = np.reshape(np.arange(255, 10*256, 256), (10,1))
dn_arange = d_arange / d_max
dn_ones = d_ones / d_max
# Test no normalization
d = (d_ones, d_arange)
r = Complex2Features._normalize_dataset((d), normalize=False)
npt.assert_array_equal(d, r)
# Test global normalization
r = Complex2Features._normalize_dataset((d_ones, ), normalize=True)
npt.assert_array_almost_equal(d_ones, r[0])
r = Complex2Features._normalize_dataset((d_arange, ), normalize=True)
npt.assert_array_almost_equal(dn_arange, r[0])
r = Complex2Features._normalize_dataset((d_ones, d_arange), normalize=True)
npt.assert_array_almost_equal(dn_ones, r[0])
npt.assert_array_almost_equal(dn_arange, r[1])
# Test one entry normalization
r = Complex2Features._normalize_dataset((d_ones, d_arange), normalize=0)
npt.assert_array_almost_equal(d_ones, r[0])
npt.assert_array_almost_equal(d_arange, r[1])
r = Complex2Features._normalize_dataset((d_ones, d_arange), normalize=1)
npt.assert_array_almost_equal(dn_ones, r[0])
npt.assert_array_almost_equal(dn_arange, r[1])
def test__data_to_dataset(self):
# Test invalid data: empty tuple
with self.assertRaises(ValueError):
Complex2Features._data_to_dataset(tuple())
# Test invalid data: shapeless type
with self.assertRaises(AttributeError):
Complex2Features._data_to_dataset(None)
# Test invalid data: different shapes
d1 = np.ones((10,256))
d2 = np.zeros((128,64))
d3 = np.zeros((10,256))
with self.assertRaises(ValueError):
Complex2Features._data_to_dataset((d1, d2))
# Test valid data: single entry
r = Complex2Features._data_to_dataset(d1)
self.assertIsInstance(r, tuple)
self.assertEqual(len(r), 1)
# Test valid data: multiple entries
r = Complex2Features._data_to_dataset((d1, d3))
self.assertIsInstance(r, tuple)
self.assertEqual(len(r), 2)
def test_c2f(self):
d = np.reshape(np.random.rand(256), (16, 16))
r1 = Complex2Features.apply(d, False, True, True, False, True)
r2 = c2f(d, False, True, True, False, True)
npt.assert_array_almost_equal(r1, r2)