def test_persistent_homology():
# Test 1 "Two bars"
max_r = 2.4
max_dim = 3
p = 5
X = np.array([[0, 0], [0, 1], [0, 2], [1, 0], [1, 2], [2, 0], [2, 0.4],
[2, 1.6], [2, 2], [3, 0], [3, 2], [4, 0], [4, 1], [4, 2]])
Dist = dist.squareform(dist.pdist(X))
compx, R = vietoris_rips(Dist, max_r, max_dim)
differentials = complex_differentials(compx, p)
Hom, Im, PreIm = persistent_homology(differentials, R, max_r, p)
cycle_bars = np.array([[1, 2], [1.2, 2]])
assert np.allclose(np.copy(Hom[1].barcode),
cycle_bars,
rtol=1e-05,
atol=1e-08)
# Test 2 "Four points"
max_r = 2
max_dim = 3
p = 5
X = np.array([[0, 0], [1, 0], [0, 1], [1, 1]])
Dist = dist.squareform(dist.pdist(X))
compx, R = vietoris_rips(Dist, max_r, max_dim)
differentials = complex_differentials(compx, p)
Hom, Im, PreIm = persistent_homology(differentials, R, max_r, p)
Im_1_res = np.array([[1., 4., 0.], [0., 1., 0.], [1., 4., 4.],
[0., 1., 1.], [0., 0., 1.], [4., 0., 0.]])
PreIm_2_res = np.array([[1., 4., 0.], [0., 1., 0.], [0., 0., 1.],
[0., 0., 0.]])
assert np.allclose(Im[1].coordinates, Im_1_res, rtol=1e-05, atol=1e-08)
assert np.allclose(PreIm[2], PreIm_2_res, rtol=1e-05, atol=1e-08)
def test_vietoris_rips():
X = np.array([[0, 0], [1, 0], [0, 1], [1, 1]])
Dist = dist.squareform(dist.pdist(X))
# Expected values
expected_complex = [
4,
np.array([[0, 1], [0, 2], [1, 3], [2, 3], [1, 2], [0, 3]]),
np.array([[0, 1, 3], [0, 2, 3], [1, 2, 3], [0, 1, 2]]),
np.array([[0, 1, 2, 3]]),
np.array([])
]
expected_R = [
np.zeros(4),
np.array([1, 1, 1, 1, np.sqrt(2), np.sqrt(2)]),
np.array([np.sqrt(2), np.sqrt(2),
np.sqrt(2), np.sqrt(2)]),
np.array([np.sqrt(2)]),
np.array([])
]
# Calculate vietoris_rips complex
viRip, R = vietoris_rips(Dist, 4, 4)
print(viRip)
assert viRip[0] == expected_complex[0]
for dim, simplices in enumerate(viRip[1:]):
assert np.array_equal(simplices, expected_complex[dim + 1])
for dim, rad in enumerate(R):
assert np.array_equal(rad, expected_R[dim])
def test_complex_differentials():
X = np.array([
[1.91580552, 0.57418571],
[-0.72993636, 1.86203999],
[1.97700111, 0.30243449],
[1.99699445, 0.10960461],
[-1.58839255, 1.21532264]])
Dist = dist.squareform(dist.pdist(X))
max_r = 3
max_dim = 4
p = 5
C, _ = vietoris_rips(Dist, max_r, max_dim)
Diff = complex_differentials(C, p)
res = np.matmul(Diff[1], Diff[2]) % p
assert np.array_equal(res, np.zeros(
(np.size(Diff[1], 0), np.size(Diff[2], 1))))
res = np.matmul(Diff[2], Diff[3]) % p
assert np.array_equal(res, np.zeros(
(np.size(Diff[2], 0), np.size(Diff[3], 1))))
def test_second_diff():
saved_data = np.loadtxt(
"test/spectral_sequence/second_differential_ex_1.txt")
no_points = int(np.size(saved_data, 0) / 3)
X = np.reshape(saved_data, (no_points, 3))
max_r = 0.39
max_dim = 4
max_div = 2
overlap = max_r
p = 5
# compute ordinary persistent homology
Dist = dist.squareform(dist.pdist(X))
C, R = vietoris_rips(Dist, max_r, max_dim)
Diff = complex_differentials(C, p)
PerHom, _, _ = persistent_homology(Diff, R, max_r, p)
# compute spectral sequence
MV_ss = create_MV_ss(X, max_r, max_dim, max_div, overlap, p)
# Check that computed barcodes coincide
for it, PH in enumerate(MV_ss.persistent_homology):
assert np.array_equal(PH.barcode, PerHom[it].barcode)
def test_second_diff():
X = np.array([
[-2.13714531e-01, 4.25223397e-01, 2.74083041e-01],
[4.01913345e-01, -4.72134869e-01, -4.73859599e-01],
[2.38914429e-01, -4.31470540e-01, 4.83776997e-01],
[-4.89370820e-01, -2.07968729e-01, -3.89636353e-01],
[4.24500295e-01, 4.58964614e-01, -4.98237341e-01],
[-4.76572637e-01, -4.06088514e-01, 4.05202146e-01],
[4.61652748e-01, 2.34792814e-01, 2.04225883e-01],
[-3.52711052e-01, 4.26715183e-01, -4.80003363e-01],
[-2.81490512e-02, -1.06009136e-01, 2.02445947e-02],
[4.96239456e-01, -4.80397512e-01, 2.17295778e-02],
[1.11144441e-01, -6.45898285e-02, -4.82885781e-01],
[1.11734990e-01, 4.44344773e-01, -1.07659800e-01],
[-5.32530741e-02, 1.75106037e-02, 4.95661087e-01],
[4.80731798e-01, -3.59756275e-02, -1.96771412e-01],
[-4.95507778e-01, -1.03466987e-02, 3.13125635e-02],
[-1.63373706e-02, -4.69398375e-01, -2.68753162e-01],
[-4.45299445e-01, -4.85796964e-01, -6.68902706e-02],
[-4.72004853e-01, 9.70583788e-02, 4.96291646e-01],
[-7.82471997e-02, -4.97384572e-01, 1.68513920e-01],
[2.00439090e-01, 4.26702152e-01, 4.68103994e-01],
[-4.70900282e-01, 4.92768146e-01, -4.51163885e-02],
[4.86542849e-01, -1.65661358e-01, 3.36148725e-01],
[-1.91356565e-01, 1.68627422e-01, -1.86292084e-01],
[5.29252474e-02, 2.94990466e-01, -4.29486145e-01],
[4.78790968e-01, 4.65502419e-01, -5.58868179e-02],
[9.46372648e-02, 1.58764102e-01, 1.97332579e-01],
[3.11750791e-01, 8.74414056e-02, 4.71358652e-01],
[-2.66839044e-01, -1.86830845e-01, 2.36774532e-01],
[2.11860827e-01, -2.41622535e-01, -1.88999919e-01],
[2.10742604e-01, -2.43169589e-01, 2.00870011e-01],
[-2.55390495e-01, -3.33269483e-03, -4.46772579e-01],
[-2.89947388e-01, -2.19377661e-01, -1.12359682e-01],
[-1.45230521e-01, -3.75468394e-01, 4.51457520e-01],
[-4.84274992e-01, 1.89424136e-01, -2.76887012e-01],
[-2.18059968e-01, 1.32276711e-01, 1.96619070e-01],
[-2.93920729e-01, -4.78747614e-01, -3.89491137e-01],
[4.94302449e-01, 2.63667032e-01, -2.82673654e-01],
[1.70233344e-01, 1.53557436e-01, -9.51354726e-02],
[4.22679639e-01, -1.79520845e-01, -4.60215694e-01],
[-4.61283014e-01, 3.95823689e-01, 4.19869086e-01],
[1.14713860e-01, 4.97274281e-01, 2.01893714e-01],
[-1.80772158e-01, 4.48634933e-01, -9.97324867e-03],
[-7.59471010e-02, 3.05404646e-01, 4.90944708e-01],
[-1.31765376e-01, -2.64189775e-01, -4.71488836e-01],
[3.50626151e-01, 8.53619684e-03, 6.31736864e-02],
[3.27342370e-01, 1.12736500e-01, -4.67830272e-01],
[2.26969648e-01, -4.80957040e-01, -5.41255144e-02],
[-3.59343891e-01, -1.74093080e-01, 4.86893652e-01],
[4.66999261e-01, -4.16984835e-01, 2.72692653e-01],
[-6.91211582e-02, -4.87917144e-02, -2.46868978e-01],
[4.34318398e-01, -3.52282399e-01, -2.30194602e-01],
[-1.85350528e-01, 4.05890330e-01, -2.62470454e-01],
[-3.90897570e-01, 2.57630258e-01, -5.62473929e-02],
[-4.88922065e-01, -8.98057939e-02, 2.88336801e-01],
[3.48242761e-02, -2.03828999e-01, 4.29397943e-01],
[1.23360188e-01, -4.05973253e-01, -4.60472677e-01],
[4.34598592e-01, 4.59320146e-01, 4.07317279e-01],
[2.76543740e-01, -1.52277597e-01, 4.94491146e-01],
[2.72507857e-01, 4.53784173e-01, -3.08576698e-01],
[-4.05842123e-02, -3.26184908e-02, 2.51198145e-01],
[-1.21954911e-03, -3.63745861e-01, -5.33832257e-02],
[-4.32475653e-01, -2.91193914e-01, 9.91502909e-02],
[1.60707196e-01, -4.96748331e-01, 1.73419212e-01],
[-2.95159529e-01, -4.48162501e-01, 2.60348915e-01],
[3.67709790e-01, 4.49301866e-01, 1.76591979e-01],
[-2.05662982e-01, -4.82046093e-01, -1.29367220e-01],
[4.35391399e-01, 1.92287067e-01, -5.30268064e-02],
[2.03230577e-01, -3.90130422e-03, -2.78116020e-01],
[-4.08775879e-02, 2.20556669e-01, 1.90769388e-02],
[-4.47807882e-01, -8.13703532e-02, -2.05568991e-01],
[-4.90648535e-01, 3.99916522e-01, 1.56957368e-01],
[2.40234266e-01, -3.81754425e-02, 2.69566353e-01],
[-1.49435190e-01, 1.81721485e-01, -4.07422016e-01],
[-1.95551049e-01, -3.59765989e-01, 5.05509993e-02],
[-3.12584419e-01, 2.36738768e-01, 4.56346884e-01],
[-2.79559496e-01, -1.41068778e-05, 3.52485705e-01],
[2.43368926e-01, -4.53663433e-01, -2.64863495e-01],
[2.50806322e-01, 3.13136489e-01, 2.98814784e-01],
[4.64167960e-01, -1.59744390e-01, 1.06879739e-01],
[4.84447427e-01, 4.69861875e-01, -2.93875544e-01],
[-4.84200093e-01, 2.13061888e-01, -4.81443616e-01],
[1.10900037e-02, -2.66281055e-01, -2.66423714e-01],
[-1.41919528e-01, 4.52894129e-01, -4.57095104e-01],
[-9.40200734e-03, 2.65552917e-01, -2.36584584e-01],
[4.35401772e-01, -4.85193017e-01, 4.90717018e-01],
[-4.55376104e-02, 3.04497817e-01, 2.07928691e-01],
[1.85612046e-02, -3.73082920e-01, 3.19525063e-01],
[2.57120177e-01, 2.21216275e-01, 1.01338061e-01],
[3.28704483e-01, -3.79758582e-01, 1.17556594e-01],
[-4.89856589e-01, 4.28481106e-01, -3.36128640e-01],
[-4.75262567e-01, -4.44026774e-01, -4.63118046e-01]])
max_r = 0.39
max_dim = 4
max_div = 2
overlap = max_r
p = 5
# compute ordinary persistent homology
Dist = dist.squareform(dist.pdist(X))
C, R = vietoris_rips(Dist, max_r, max_dim)
Diff = complex_differentials(C, p)
PerHom, _, _ = persistent_homology(Diff, R, max_r, p)
# compute spectral sequence
MV_ss = create_MV_ss(X, max_r, max_dim, max_div, overlap, p)
# Check that computed barcodes coincide
for it, PH in enumerate(MV_ss.persistent_homology):
assert np.array_equal(PH.barcode, PerHom[it].barcode)
def test_double_complex():
# creating and saving new point cloud ############
# X = random_cube(1000, 3)
# point_cloud = take_sample(X, 180)
# output_file = open("test/spectral_sequence/random_cube.txt", "w")
# for row in point_cloud:
# np.savetxt(output_file, row)
# output_file.close()
# using old point cloud ###################
saved_data = np.loadtxt("test/spectral_sequence/random_cube.txt")
no_points = int(np.size(saved_data, 0) / 3)
point_cloud = np.reshape(saved_data, (no_points, 3))
max_r = 0.2
max_dim = 3
max_div = 3
overlap = max_r * 1.01
p = 3
# compute ordinary persistent homology
Dist = dist.squareform(dist.pdist(point_cloud))
C, R = vietoris_rips(Dist, max_r, max_dim)
Diff = complex_differentials(C, p)
PerHom, _, _ = persistent_homology(Diff, R, max_r, p)
###########################################################################
# compute spectral sequence
MV_ss = create_MV_ss(point_cloud, max_r, max_dim, max_div, overlap, p)
###########################################################################
# check that first page representatives are cycles
for n_deg in range(MV_ss.no_columns):
if n_deg == 0:
no_covers = MV_ss.nerve[0]
else:
no_covers = len(MV_ss.nerve[n_deg])
for nerve_spx_index in range(no_covers):
local_differentials = complex_differentials(
MV_ss.subcomplexes[n_deg][nerve_spx_index], p)
for mdeg, hom in enumerate(
MV_ss.Hom[0][n_deg][nerve_spx_index][1:]):
if hom.dim > 0:
trivial_image = np.matmul(local_differentials[mdeg + 1],
hom.coordinates)
if np.any(trivial_image % p):
raise (RuntimeError)
# end if
# end if
# end for
# end for
# end for
###########################################################################
# compute cech differential twice and check that it vanishes
"""Iterate over test_local_cech_matrix"""
for n_dim in range(2, MV_ss.no_columns):
for deg in range(MV_ss.no_rows):
if MV_ss.page_dim_matrix[1][deg][n_dim] > 0:
cech_differential_twice(MV_ss, n_dim, deg)
# end if
# end for
# end for
###########################################################################
# check that classic PH coincides with result
for it, PH in enumerate(MV_ss.persistent_homology):
assert np.array_equal(PH.barcode, PerHom[it].barcode)
def test_double_complex():
X = random_cube(500, 3)
point_cloud = take_sample(X, 50)
max_r = 0.4
max_dim = 3
max_div = 2
overlap = max_r * 1.01
p = 3
# compute ordinary persistent homology
Dist = dist.squareform(dist.pdist(point_cloud))
C, R = vietoris_rips(Dist, max_r, max_dim)
Diff = complex_differentials(C, p)
PerHom, _, _ = persistent_homology(Diff, R, max_r, p)
# compute spectral sequence
MV_ss = create_MV_ss(point_cloud, max_r, max_dim, max_div, overlap, p)
# Check that computed barcodes coincide
for it, PH in enumerate(MV_ss.persistent_homology):
assert np.array_equal(PH.barcode, PerHom[it].barcode)
# Vertical Differentials
for n_dim in range(MV_ss.no_columns):
# check for vertical differential
if n_dim == 0:
rk_nerve = MV_ss.nerve[n_dim]
else:
rk_nerve = len(MV_ss.nerve[n_dim])
for nerv_spx in range(rk_nerve):
for deg in range(2, MV_ss.no_rows):
zero_coordinates = []
coord_matrix = np.identity(
len(MV_ss.subcomplexes[n_dim][nerv_spx][deg]))
for spx_idx in range(
len(MV_ss.subcomplexes[n_dim][nerv_spx][deg])):
zero_coordinates.append({nerv_spx: coord_matrix[spx_idx]})
# end for
im_vert = MV_ss.vert_image(zero_coordinates, n_dim, deg)
im_vert = MV_ss.vert_image(im_vert, n_dim, deg - 1)
for i, coord in enumerate(im_vert):
for n in iter(coord):
if np.any(coord[n]):
assert False
# end if
# end for
# end for
# end for
# end for
# end for
# check for Cech differentials
for deg in range(MV_ss.no_rows):
for n_dim in range(2, MV_ss.no_columns):
for nerv_spx in range(len(MV_ss.nerve[n_dim])):
zero_coordinates = []
if deg == 0:
no_simplexes = MV_ss.subcomplexes[n_dim][nerv_spx][deg]
else:
no_simplexes = len(
MV_ss.subcomplexes[n_dim][nerv_spx][deg])
coord_matrix = np.identity(no_simplexes)
for spx_idx in range(no_simplexes):
zero_coordinates.append({nerv_spx: coord_matrix[spx_idx]})
# end for
cech_diff = MV_ss.cech_differential(zero_coordinates, n_dim,
deg)
cech_diff = MV_ss.cech_differential(cech_diff, n_dim - 1, deg)
for i, coord in enumerate(cech_diff):
for n in iter(coord):
if np.any(coord[n]):
assert False
# end if
# end for
# end for
# end for
# end for
# end for
# check for anticommutativity
for n_dim in range(1, MV_ss.no_columns):
for nerv_spx in range(len(MV_ss.nerve[n_dim])):
for deg in range(1, MV_ss.no_rows):
zero_coordinates = []
coord_matrix = np.identity(
len(MV_ss.subcomplexes[n_dim][nerv_spx][deg]))
for spx_idx in range(
len(MV_ss.subcomplexes[n_dim][nerv_spx][deg])):
zero_coordinates.append({nerv_spx: coord_matrix[spx_idx]})
# end for
im_left = MV_ss.cech_differential(zero_coordinates, n_dim, deg)
im_left = MV_ss.vert_image(im_left, n_dim - 1, deg)
im_right = MV_ss.vert_image(zero_coordinates, n_dim, deg)
im_right = MV_ss.cech_differential(im_right, n_dim, deg - 1)
for i, coord in enumerate(im_left):
result = add_dictionaries([1, 1], [coord, im_right[i]],
MV_ss.p)
for n in iter(result):
if np.any(result[n]):
assert False
def test_zig_zags():
# creating and saving new point cloud ############
X = torus3D(1000, 1, 3)
point_cloud = take_sample(X, 300)
output_file = open("test/spectral_sequence/torus3D.txt", "w")
for row in point_cloud:
np.savetxt(output_file, row)
output_file.close()
# using old point cloud ###################
# saved_data = np.loadtxt("test/spectral_sequence/failing_1.txt")
# no_points = int(np.size(saved_data,0) / 3)
# point_cloud = np.reshape(saved_data, (no_points, 3))
max_r = 1
max_dim = 3
max_div = 2
overlap = max_r * 1.01
p = 3
# compute ordinary persistent homology
Dist = dist.squareform(dist.pdist(point_cloud))
C, R = vietoris_rips(Dist, max_r, max_dim)
Diff = complex_differentials(C, p)
PerHom, _, _ = persistent_homology(Diff, R, max_r, p)
###########################################################################
# compute spectral sequence
MV_ss = create_MV_ss(point_cloud, max_r, max_dim, max_div, overlap, p)
###########################################################################
# check that first page representatives are cycles
for n_deg in range(MV_ss.no_columns):
if n_deg == 0:
no_covers = MV_ss.nerve[0]
else:
no_covers = len(MV_ss.nerve[n_deg])
for nerve_spx_index in range(no_covers):
for mdeg, hom in enumerate(
MV_ss.Hom[0][n_deg][nerve_spx_index][1:]):
if hom.dim > 0:
trivial_image = np.matmul(
MV_ss.zero_diff[n_deg][nerve_spx_index][mdeg + 1],
hom.coordinates)
if np.any(trivial_image % p):
raise (RuntimeError)
# end if
# end if
# end for
# end for
# end for
# TEST commutativity of zig-zags
########################################################################
for start_n_dim in range(1, MV_ss.no_columns):
for start_deg in range(MV_ss.no_rows):
Sn_dim, Sdeg = start_n_dim, start_deg
for k, chains in enumerate(
MV_ss.Hom_reps[MV_ss.no_pages -
1][start_n_dim][start_deg][:-1]):
# calculate Cech differential of chains in (Sn_dim, Sdeg)
diff_im = MV_ss.cech_diff(Sn_dim - 1, Sdeg, chains)
# calculate vertical differential of (Sn_dim - 1, Sdeg + 1)
Sn_dim, Sdeg = Sn_dim - 1, Sdeg + 1
next_chains = MV_ss.Hom_reps[MV_ss.no_pages -
1][start_n_dim][start_deg][k + 1]
vert_im = local_chains(len(next_chains.ref))
for nerve_spx_index, ref in enumerate(next_chains.ref):
if len(ref) > 0:
# compute vertical image of next_chains locally
vert_im.add_entry(
nerve_spx_index, ref,
np.matmul(
MV_ss.zero_diff[Sn_dim][nerve_spx_index][Sdeg],
next_chains.coord[nerve_spx_index].T).T % p)
# end if
# end for
# check that d_v(next_chains) + d_cech(chains) = 0
zero_chains = vert_im + diff_im
for coord in zero_chains.coord:
if len(coord) > 0:
if np.any(coord % p):
raise (ValueError)