def test8():
"""Same as test1, but now comparing different autograd methods."""
T, n = projective_mds.circleRPn()
D = projective_mds.graph_distance_matrix(T, k=5)
Y = projective_mds.initial_guess(T, 2)
costs, times = do_autograd_tests(Y, D)
return costs, times
def test10(solve_prog='autograd'):
"""Test using the bezier curve on RP^4 (which gets tangled by PPCA)."""
B = np.load('bez_test.npy')
D = projective_mds.graph_distance_matrix(B, k=5)
Y = projective_mds.initial_guess(B, 2)
out = do_autograd_tests(Y, D, solve_prog=solve_prog)
return out
def test4():
Ws = np.load('workspace.npz')
T = Ws['BB']
Y = Ws['Y']
D = projective_mds.graph_distance_matrix(T, k=8)
out = do_tests(Y, D)
return out
def test9():
"""Same as test4, but with autograd methods."""
Ws = np.load('workspace.npz')
T = Ws['BB']
Y = Ws['Y']
D = projective_mds.graph_distance_matrix(T, k=8)
costs, times = do_autograd_tests(Y, D)
return costs, times
def test12(pmo_solve='cg'):
"""Same as test4, but with autograd methods."""
Ws = np.load('workspace.npz')
T = Ws['BB']
Y = Ws['Y']
D = projective_mds.graph_distance_matrix(T, k=8)
out = compare_gradients(Y, D, pmo_solve=pmo_solve)
return out
def test5():
T, _ = projective_mds.circleRPn(noise=True, segment_points=60)
D = projective_mds.graph_distance_matrix(T, k=8)
Y, _ = projective_mds.circleRPn(dim=2,
segment_points=100,
num_segments=2,
noise=True,
v=0.1)
out = do_tests(Y, D)
return out
def test7():
"""Test applying PMDS before PPCA, with the tangled curve."""
B = np.load('bez_test.npy')
D = projective_mds.graph_distance_matrix(B, k=5)
X, C = do_tests(B, D, plot=False)
Y1 = projective_mds.initial_guess(X[0], 2)
Y2 = projective_mds.initial_guess(X[1], 2)
Y3 = projective_mds.initial_guess(X[2], 2)
Y4 = projective_mds.initial_guess(X[3], 2)
Y5 = projective_mds.initial_guess(X[4], 2)
out = (Y1, Y2, Y3, Y4, Y5)
return out
def test1():
T, n = projective_mds.circleRPn()
D = projective_mds.graph_distance_matrix(T, k=5)
Y = projective_mds.initial_guess(T, 2)
out = do_tests(Y, D)
return out
def test11(pmo_solve='cg'):
B = np.load('bez_test.npy')
D = projective_mds.graph_distance_matrix(B, k=5)
Y = projective_mds.initial_guess(B, 2)
out = compare_gradients(Y, D, pmo_solve=pmo_solve)
return out
