def test_optimisation(load_magnetic_anomaly):
d = load_magnetic_anomaly["mag_data"]
xc = load_magnetic_anomaly["xc"]
yc = load_magnetic_anomaly["yc"]
xmin, xmax, ymin, ymax = load_magnetic_anomaly["extent"]
max_window = load_magnetic_anomaly["max_window"]
grid = pycurious.CurieOptimise(d, xmin, xmax, ymin, ymax)
beta, zt, dz, C = grid.optimise(max_window, xc, yc, taper=np.hanning)
x_opt = np.array([beta, zt, dz])
# hard-coded parameters used to generate the magnetic anomaly
zt0 = 0.305
dz0 = 10.0 + zt0
beta0 = 3.0
x0 = np.array([beta0, zt0, dz0])
# compare if they are close or not
# some parameters should be more similar than others
tol = np.array([0.3, 0.1, 2.0])
parameters = ["beta", "zt", "dz"]
err_msg = "FAILED! {} = {:.4f} is not within an acceptable tolerance of {}"
for i in range(x0.size):
npt.assert_allclose(
x_opt[i],
x0[i],
atol=tol[i],
err_msg=err_msg.format(parameters[i], x_opt[i], tol[i]),
)
def test_priors():
grid = pycurious.CurieOptimise(mag_grid, xmin, xmax, ymin, ymax)
beta0, zt0, dz0, C0 = grid.optimise(max_window, xc, yc)
grid.add_prior(beta=(1.0, 0.1))
beta1, zt1, dz1, C1 = grid.optimise(max_window, xc, yc)
if abs(beta1 - 1.0) < abs(beta0 - 1.0):
print("PASSED! Optimisation is sensitive to prior information")
else:
assert False, "FAILED! Optimisation with priors failed"
def test_valid_numbers():
S = np.array([ 22.16409774, 19.95258494, 18.27873722, 17.10575637,\
16.53959747, 16.31539575, 15.69619005, 15.29953307,\
14.83475976, 14.54031396, 14.33361716, 13.81764026,\
13.5176055 , 13.27386563, 13.03493328, 12.88581369,\
12.61998358, 12.48616749, 12.11261083, 12.13079154,\
11.85440661, 11.79244826, 11.66823202, 11.40231744,\
11.32521296, 11.13634007, 11.10650999, 10.94822598,\
10.78032794, 10.66593304, 10.55815845, 10.56805594,\
10.33514462, 10.22026537, 10.22945756, 10.09275259,\
10.11562101, 9.85061009, 9.87165772, 9.85976847,\
9.73954992, 9.72021054, 9.52959744, 9.59582531,\
9.50927273, 9.44691364, 9.39293966, 9.33097387, 9.33191784])
k = np.array([ 0.09237068, 0.15443902, 0.21486191, 0.282031 , 0.33452175,\
0.4020083 , 0.46381582, 0.5290359 , 0.59611689, 0.65875047,\
0.71442694, 0.77929408, 0.84320843, 0.90939865, 0.96863721,\
1.03155349, 1.09130906, 1.15699532, 1.22338685, 1.28109263,\
1.34606572, 1.40861721, 1.46920824, 1.53421046, 1.59675833,\
1.66109235, 1.7233503 , 1.78492941, 1.84835837, 1.90933735,\
1.973275 , 2.03650008, 2.0983359 , 2.16386249, 2.22416603,\
2.2876692 , 2.34841931, 2.41225232, 2.47774604, 2.5379226 ,\
2.59980401, 2.66346692, 2.72735801, 2.79146076, 2.85220715,\
2.91443307, 2.97643404, 3.04008782, 3.10369815])
sigma_S = np.ones_like(S)
grid = pycurious.CurieOptimise(mag_grid, xmin, xmax, ymin, ymax)
beta0 = 3.0
zt0 = 1.0
dz0 = 40.0
C0 = 5.0
x0 = np.array([beta0, zt0, dz0, C0])
lower_bound = np.zeros_like(x0)
upper_bound = [None] * len(lower_bound)
# xi = func([5, 0., 45., 6.], S, k)
xi = grid.min_func(x0, k, S, sigma_S)
res = minimize(grid.min_func,
x0,
args=(k, S, sigma_S),
method='TNC',
bounds=list(zip(lower_bound, upper_bound)))
print("beta={:.2f}, zt={:.2f}, dz={:.2f}, C={:.2f}".format(*res.x))
if np.isfinite(res.x).all():
print("PASSED! All numbers are finite")
else:
assert False, "FAILED! Not all numbers are finite"
def test_priors(load_magnetic_anomaly):
d = load_magnetic_anomaly["mag_data"]
xc = load_magnetic_anomaly["xc"]
yc = load_magnetic_anomaly["yc"]
xmin, xmax, ymin, ymax = load_magnetic_anomaly["extent"]
max_window = load_magnetic_anomaly["max_window"]
grid = pycurious.CurieOptimise(d, xmin, xmax, ymin, ymax)
beta0, zt0, dz0, C0 = grid.optimise(max_window, xc, yc)
grid.add_prior(beta=(1.0, 0.1))
beta1, zt1, dz1, C1 = grid.optimise(max_window, xc, yc)
assert abs(beta1 -
1.0) < abs(beta0 -
1.0), "FAILED! Optimisation with priors failed"
def test_CurieOptimise_routines(load_magnetic_anomaly):
d = load_magnetic_anomaly["mag_data"]
xc = load_magnetic_anomaly["xc"]
yc = load_magnetic_anomaly["yc"]
xmin, xmax, ymin, ymax = load_magnetic_anomaly["extent"]
max_window = load_magnetic_anomaly["max_window"]
cpd = pycurious.CurieOptimise(d, xmin, xmax, ymin, ymax)
xc_list, yc_list = cpd.create_centroid_list(0.5 * max_window,
spacingX=5e3,
spacingY=5e3)
xc = xc_list[0]
yc = yc_list[0]
assert time_routine(cpd.add_prior, beta=(3.0, 0.5))
assert time_routine(cpd.optimise, 0.5 * max_window, xc, yc)
assert time_routine(cpd.optimise_routine, 0.5 * max_window, xc_list,
yc_list)
assert time_routine(cpd.metropolis_hastings, 0.5 * max_window, xc, yc, 100,
10)
assert time_routine(cpd.sensitivity, 0.5 * max_window, xc, yc, 100)
max_window = 100e3
# square grid 100km x 100km
xmin = 0.0
xmax = 100e3
ymin = 0.0
ymax = 100e3
nx = 100
ny = 100
# random noise
mag_grid = np.random.randn(ny, nx)
cpd = pycurious.CurieOptimise(mag_grid, xmin, xmax, ymin, ymax)
# compute centroid
xc = 0.5 * (xmin + xmax)
yc = 0.5 * (ymin + ymax)
xc_list, yc_list = cpd.create_centroid_list(0.5 * max_window,
spacingX=5e3,
spacingY=5e3)
print("number of centroids = {}".format(len(xc_list)))
t = time.time()
betaN, ztN, dzN, CN = cpd.optimise_routine(0.5 * max_window,
xc_list,
yc_list,
taper=None)
import pycurious
# load x,y,anomaly
mag_data = np.loadtxt("../pycurious/Examples/data/test_mag_data.txt")
nx, ny = 305, 305
x = mag_data[:,0]
y = mag_data[:,1]
d = mag_data[:,2].reshape(ny,nx)
xmin, xmax = x.min(), x.max()
ymin, ymax = y.min(), y.max()
# initialise CurieOptimise object
grid = pycurious.CurieOptimise(d, xmin, xmax, ymin, ymax)
# get centroids
window_size = 200e3
xc_list, yc_list = grid.create_centroid_list(window_size, spacingX=10e3, spacingY=10e3)
print("number of centroids = {}".format(len(xc_list)))
beta, zt, dz, C = grid.optimise_routine(window_size, xc_list, yc_list)
print "done with optimise_routine"
result = grid.parallelise_routine(window_size, xc_list, yc_list, grid.sensitivity, 5)
print result
result1 = np.array(result)
def test_valid_numbers(load_magnetic_anomaly):
d = load_magnetic_anomaly["mag_data"]
xc = load_magnetic_anomaly["xc"]
yc = load_magnetic_anomaly["yc"]
xmin, xmax, ymin, ymax = load_magnetic_anomaly["extent"]
max_window = load_magnetic_anomaly["max_window"]
# create phoney power spectrum
S = np.array([
22.16409774,
19.95258494,
18.27873722,
17.10575637,
16.53959747,
16.31539575,
15.69619005,
15.29953307,
14.83475976,
14.54031396,
14.33361716,
13.81764026,
13.5176055,
13.27386563,
13.03493328,
12.88581369,
12.61998358,
12.48616749,
12.11261083,
12.13079154,
11.85440661,
11.79244826,
11.66823202,
11.40231744,
11.32521296,
11.13634007,
11.10650999,
10.94822598,
10.78032794,
10.66593304,
10.55815845,
10.56805594,
10.33514462,
10.22026537,
10.22945756,
10.09275259,
10.11562101,
9.85061009,
9.87165772,
9.85976847,
9.73954992,
9.72021054,
9.52959744,
9.59582531,
9.50927273,
9.44691364,
9.39293966,
9.33097387,
9.33191784,
])
k = np.array([
0.09237068,
0.15443902,
0.21486191,
0.282031,
0.33452175,
0.4020083,
0.46381582,
0.5290359,
0.59611689,
0.65875047,
0.71442694,
0.77929408,
0.84320843,
0.90939865,
0.96863721,
1.03155349,
1.09130906,
1.15699532,
1.22338685,
1.28109263,
1.34606572,
1.40861721,
1.46920824,
1.53421046,
1.59675833,
1.66109235,
1.7233503,
1.78492941,
1.84835837,
1.90933735,
1.973275,
2.03650008,
2.0983359,
2.16386249,
2.22416603,
2.2876692,
2.34841931,
2.41225232,
2.47774604,
2.5379226,
2.59980401,
2.66346692,
2.72735801,
2.79146076,
2.85220715,
2.91443307,
2.97643404,
3.04008782,
3.10369815,
])
sigma_S = np.ones_like(S)
grid = pycurious.CurieOptimise(d, xmin, xmax, ymin, ymax)
beta0 = 3.0
zt0 = 1.0
dz0 = 40.0
C0 = 5.0
x0 = np.array([beta0, zt0, dz0, C0])
lower_bound = np.zeros_like(x0)
upper_bound = [None] * len(lower_bound)
# xi = func([5, 0., 45., 6.], S, k)
xi = grid.min_func(x0, k, S, sigma_S)
res = minimize(
grid.min_func,
x0,
args=(k, S, sigma_S),
method="TNC",
bounds=list(zip(lower_bound, upper_bound)),
)
print("beta={:.2f}, zt={:.2f}, dz={:.2f}, C={:.2f}".format(*res.x))
parameters = ["beta", "zt", "dz", "C"]
err_msg = "FAILED! {} = {} is not a finite number"
for i in range(res.x.size):
assert np.isfinite(res.x[i]), err_msg.format(parameters[i], res.x[i])
def test_optimisation():
grid = pycurious.CurieOptimise(mag_grid, xmin, xmax, ymin, ymax)
beta, zt, dz, C = grid.optimise(max_window, xc, yc)