def fit(self):
"""
fit scan to an appropriate spline
calculate a set of partition function values at different temperatures and fit a spline to the
partition function values
generate splines for the first and second derivatives of the partition function
"""
N = len(self.pivots)
if N > 2:
self.V = interpolate.LinearNDInterpolator(self.phis, self.Es)
self.rootD = interpolate.LinearNDInterpolator(
self.phis, self.rootDs)
elif N == 2:
self.V = interpolate.SmoothBivariateSpline(self.phis[:, 0],
self.phis[:,
1], self.Es)
self.rootD = interpolate.SmoothBivariateSpline(
self.phis[:, 0], self.phis[:, 1], self.rootDs)
else:
self.V = interpolate.CubicSpline(self.phis, self.Es)
self.rootD = interpolate.CubicSpline(self.phis, self.rootDs)
Tlist = np.linspace(10.0, 3001.0, num=20, dtype=np.float64)
Qs = []
for T in Tlist:
Qs.append(self.calc_partition_function(T))
self.Q = interpolate.CubicSpline(Tlist, Qs)
self.dQdT = self.Q.derivative()
self.d2QdT2 = self.dQdT.derivative()
def find_spline_transform(a_t, sm, tm):
"""
Determine the residual `x` and `y` offsets between matched coordinates
after affine transformation and fit 2D spline surfaces to describe the
spatially-varying correction to be applied.
"""
spline_order = 3
# Get the source, after affine transformation, and template coordinates
source_coo = a_t.apply_transform(get_det_coords(sm))
template_coo = get_det_coords(tm)
# Create splines describing the residual offsets in x and y left over
# after the affine transformation
kx = ky = spline_order
sbs_x = interpolate.SmoothBivariateSpline(
template_coo[:, 0],
template_coo[:, 1], (template_coo[:, 0] - source_coo[:, 0]),
kx=kx,
ky=ky)
sbs_y = interpolate.SmoothBivariateSpline(
template_coo[:, 0],
template_coo[:, 1], (template_coo[:, 1] - source_coo[:, 1]),
kx=kx,
ky=ky)
return (sbs_x, sbs_y)
def scatter2smoothgridded(outputfile,
headerfile=None,
xi=[],
yi=[],
inputfile=None,
x=[],
y=[],
z=[],
topotype=2,
kx=3,
ky=3,
s=0):
"""
interpolate scattered data to gridded data using a smoothed bivariate spline of order kx,ky
note: this will heavily smooth a complicated DEM but can be of use for
new grid defined by uniformly spaced xi, yi or a header file
input data can be x,y,z arrays or an .xyz 3 column file
"""
from scipy import interpolate
if (not x) | (not y) | (not z):
try:
inputfile
except NameError:
err_msg = "Error: you must call with either a data file or x,y,z arrays"
raise Exception(err_msg)
if (not xi) | (not yi):
try:
headerfile
except NameError:
err_msg = "Error: you must call with either a header file or xi,yi coordinate arrays"
raise Exception(err_msg)
if inputfile:
xyz = np.loadtxt(inputfile)
x = xyz[:, 0]
y = xyz[:, 1]
z = xyz[:, 2]
if headerfile:
#determine what the output grid will look like from headerfile
topoheader = topoheaderread(inputfile=headerfile)
#Create the gridded data using pylab "griddata function."
xi = np.arange(topoheader['xll'], \
topoheader['xll']+topoheader['ncols']*topoheader['cellsize'], \
step=topoheader['cellsize'],dtype=float)
yi = np.arange(topoheader['yll'], \
topoheader['yll']+topoheader['nrows']*topoheader['cellsize'], \
step=topoheader['cellsize'],dtype=float)
w = np.ones(np.shape(x)) / np.shape(x)
sp = interpolate.SmoothBivariateSpline(y, x, z, w=w, kx=kx, ky=ky, s=s)
Zi = sp(yi, xi)
(Xi, Yi) = np.meshgrid(xi, yi)
Yi = np.flipud(Yi)
Zi = np.flipud(Zi)
griddata2topofile(Xi, Yi, Zi, outputfile, topotype)
def fine_dewarp(im, lines):
im_h, im_w = im.shape[:2]
debug = cv2.cvtColor(im, cv2.COLOR_GRAY2BGR)
points = []
y_offsets = []
for line in lines:
if len(line) < 10 or abs(line.fit_line().angle()) > 0.001: continue
line.fit_line().draw(debug, thickness=1)
base_points = np.array([letter.base_point() for letter in line.inliers()])
median_y = np.median(base_points[:, 1])
y_offsets.append(median_y - base_points[:, 1])
points.append(base_points)
for underline in line.underlines:
mid_contour = (underline.top_contour() + underline.bottom_contour()) / 2
all_mid_points = np.stack([
underline.x + np.arange(underline.w), mid_contour,
])
mid_points = all_mid_points[:, ::4]
points.append(mid_points)
for p in base_points:
pt = tuple(np.round(p).astype(int))
cv2.circle(debug, (pt[0], int(median_y)), 2, lib.RED, -1)
cv2.circle(debug, pt, 2, lib.GREEN, -1)
cv2.imwrite('points.png', debug)
points = np.concatenate(points)
y_offsets = np.concatenate(y_offsets)
mesh = np.mgrid[:im_w, :im_h].astype(np.float32)
xmesh, ymesh = mesh
# y_offset_interp = interpolate.griddata(points, y_offsets, xmesh, ymesh, method='nearest')
# y_offset_interp = y_offset_interp.clip(-5, 5)
# mesh[1] += y_offset_interp # (mesh[0], mesh[1], grid=False)
y_offset_interp = interpolate.SmoothBivariateSpline(
points[:, 0], points[:, 1], y_offsets.clip(-3, 3),
s=4 * points.shape[0]
)
ymesh -= y_offset_interp(xmesh, ymesh, grid=False).clip(-3, 3)
conv_xmesh, conv_ymesh = cv2.convertMaps(xmesh, ymesh, cv2.CV_16SC2)
out = cv2.remap(im, conv_xmesh, conv_ymesh,
interpolation=cv2.INTER_LINEAR,
borderValue=np.median(im)).T
cv2.imwrite('corrected.png', out)
debug = cv2.cvtColor(out, cv2.COLOR_GRAY2BGR)
for line in lines:
base_points = np.array([letter.base_point() for letter in line.inliers()[1:-1]])
base_points[:, 1] -= y_offset_interp(base_points[:, 0], base_points[:, 1], grid=False)
Line.fit(base_points).draw(debug, thickness=1)
cv2.imwrite('corrected_line.png', debug)
return out
def get_interpolator(_pts):
""" Create bivariate spline interpolator """
x = [pt.x for pt in _pts]
y = [pt.y for pt in _pts]
z = [pt.z for pt in _pts]
_interp_f = interpolate.SmoothBivariateSpline(x, z, y, kx=4, ky=4)
return _interp_f
def fit_TPpwvEL_curve(self, pwv_vector, EL_vector):
"""
Fits a curve that relates the elevation EL and the KID power Pkid to the
sky temperature Tb_sky. A smooth bivariate spline or third order is used
for the interpolation. A separate 2D function is made for each filter in
the filterbank of the MKID chip and each function is saved in a separate
file.
Parameters
------------
EL_vector: vector or scalar
Values of the elevation for which the KID power and sky temperature
are to be calculated.
Unit: degrees
pwv: vector or scalar
Values of the precipitable water vapor for which the KID power and
sky temperature are to be calculated.
Unit: mm
"""
length_EL_vector = len(EL_vector)
# eta_atm, F = self.load_etaF_data()
# peak_indices = find_peaks(eta_atm[0, :]*(-1))[0] #gives indices of peaks
#obtain data
Tb_sky, Pkid = self.load_TP_data(pwv_vector, EL_vector)
# make vectors of matrices
pwv_vector_long = np.array([])
EL_vector_long = np.array([])
for i in range(0, len(pwv_vector)):
pwv_vector_long = np.append(pwv_vector_long, pwv_vector[i]*np.ones(length_EL_vector))
EL_vector_long = np.append(EL_vector_long, EL_vector)
# make interpolations
for j in range(0, self.num_filters):
split_Tb_sky = tuple(np.vsplit(Tb_sky[:, j, :], len(Tb_sky[:, 0])))
Tb_sky_vector = np.hstack(split_Tb_sky)
split_Pkid = tuple(np.vsplit(Pkid[:, j, :], len(Pkid[:, 0])))
Pkid_vector = np.hstack(split_Pkid)
# if j in peak_indices:
EL_vector_long = EL_vector_long.reshape([1, EL_vector_long.size])
f = interpolate.SmoothBivariateSpline(EL_vector_long, Pkid_vector, \
Tb_sky_vector, s = len(EL_vector_long))
# f_pwv = interpolate.SmoothBivariateSpline(Pkid_vector, EL_vector_long, \
# pwv_vector_long, s = len(Pkid_vector), kx = 3, ky = 3)
if self.D1:
name = self.path_model + '\Data\splines_Tb_sky\spline_' + '%.1f' % (self.filters[j]/1e9) +'GHz_D1'
else:
name = self.path_model + '\Data\splines_Tb_sky\spline_' + '%.1f' % (self.filters[j]/1e9) +'GHz'
# name_pwv = self.path_model + '\Data\splines_pwv\spline_' + '%.1f' % (self.filters[j]/1e9) +'GHz_D1'
np.save(name, np.array(f))
f_load = np.load(name + '.npy', allow_pickle= True)
f_function = f_load.item()
return 0
def make_spline(x, y, z, interpolator, x_grid, y_grid, mask):
'''
Convenient wrapper to abstract away interpolation.
inputs: x: numpy array of x values
y: " " y " "
z: " " z " "
interpolator: string of either "radial", "nearest", or "smooth" to choose interpolating
function.
x_grid: mesh of x values over which to interpolate
y_grid: " " y " "
mask: boolean mask of areas to exclude from the x,y mesh
returns: mesh grid, inputs x,y,z after filtering, and the filtered z mesh values Z
'''
# If there is a nan temperature, we can't plot it so we remove it
temp_mask = ~np.isnan(z)
z = z[temp_mask]
x = x[temp_mask]
y = y[temp_mask]
pos_mask = np.where(y < -70)
z = z[pos_mask]
x = x[pos_mask]
y = y[pos_mask]
B1, B2 = np.meshgrid(x_grid, y_grid, indexing='xy')
points = np.array([x, y]).T
values = z
if interpolator == 'radial':
spline = sp.interpolate.Rbf(x, y, z, function='cubic', smooth=5)
Z = spline(B1.T, B2.T)
if interpolator == 'nearest':
spline = it.NearestNDInterpolator(points, values)
Z = spline(B1.T, B2.T)
if interpolator == 'smooth':
spline = it.SmoothBivariateSpline(x, y, z)
Z = spline.ev(B1.T, B2.T)
for i in range(len(B2)):
for j in range(len(B2)):
if mask[i][j]:
continue
else:
Z[i][j] = None
return B1, B2, Z, x, y, z
def check_backgrounds():
"""
UDF
Plot the automatically-determined backgrounds as a function of position
in the UDF frame
"""
c = catIO.Readfile('../F140W/HUDF12-F140W.reform.cat')
ok = c.mag_auto < 27.5
fp = open('udf_backgrounds.dat','w')
fp.write('# id c0 cx cy x y mag\n')
for i in np.arange(c.N)[ok]:
id = c.number[i]
bgfile = 'UDF_FIT/UDF_%05d.bg.dat' %(id)
if os.path.exists(bgfile):
line = open(bgfile).readlines()[1][:-1]
fp.write('%s %7.1f %7.1f %.2f\n' %(line, c.x_image[i], c.y_image[i], c.mag_auto[i]))
fp.close()
bg = catIO.Readfile('udf_backgrounds.dat')
ok = (bg.mag > 24) & (bg.c0+0.003 > 0) & (bg.c0+0.003 < 0.004)
plt.scatter(bg.x[ok], bg.y[ok], c=bg.c0[ok], s=30, vmin=-0.004, vmax=0.002)
#### Try 2D as in sciypy example (need newer version)
# from scipy.interpolate import griddata
#
# points = [bg.x[ok], bg.y[ok]]
# values = bg.c0[ok]
#
# grid_z2 = griddata(points, values, (bg.x, bg.y), method='cubic')
from scipy import interpolate
bg_spline = interpolate.SmoothBivariateSpline(bg.x[ok], bg.y[ok], bg.c0[ok], kx=4, ky=4)
test = bg.c0*0.
for i in range(bg.N):
test[i] = bg_spline(bg.x[i], bg.y[i])
#
plt.scatter(bg.x+3600, bg.y, c=test, s=30, vmin=-0.004, vmax=0.002)
plt.text(2000,3800,'Data (mag > 24)', ha='center', va='center')
plt.text(5600,3800,'Interpolated', ha='center', va='center')
plt.savefig('background_spline.pdf')
import cPickle as pickle
fp = open('background_spline.pkl','wb')
pickle.dump(bg_spline, fp)
fp.close()
def offset_base(self, xyin):
"""Derive Displacement at the zenith
Parameters
----------
xyin : `np.ndarray`, (N, 2)
Input coordinates.
Unit is degree for sky, mm for PFI, and pixel for MCS.
Returns
-------
offsetx : `np.ndarray`, (N, 1)
Displacement in x-axis.
offsety : `np.ndarray`, (N, 1)
Displacement in y-axis.
"""
if self.skip1_off:
logging.info("------ Skipped.")
offsetx = offsety = np.zeros(xyin.shape[1])
else:
# sky-x sky-y off-x off-y
dfile = mypath+"data/offset_base_"+self.mode+".dat"
IpolD = np.loadtxt(dfile).T
x_itrp = ipol.SmoothBivariateSpline(IpolD[0, :], IpolD[1, :],
IpolD[2, :], kx=5, ky=5, s=1)
y_itrp = ipol.SmoothBivariateSpline(IpolD[0, :], IpolD[1, :],
IpolD[3, :], kx=5, ky=5, s=1)
logging.info("Interpolated the base offset")
offsetx = np.array([x_itrp.ev(i, j) for i, j in zip(*xyin)])
offsety = np.array([y_itrp.ev(i, j) for i, j in zip(*xyin)])
return offsetx, offsety
def get_interpolated_pstrain():
[x, y, pstrain] = get_geodynamic_pstrain()
# 'fdfault' generated grid (curvilinear)
result = fdfault.output('longterm_initial_mesh', 'vxbody')
result.load()
# added_layer_thickness = 60.
# print(nby1, nby0, nby0+nby1)
# plt.pcolor(result.x[:, nby0:nby0+nby1], result.y[:, nby0:nby0+nby1], result.vx[:, nby0:nby0+nby1])
# plt.scatter(x,y)
# plt.show()
# print(np.shape(result.x))
# print(np.shape(result.y[:,0:nby1]))
# print(result.y[0,0:nby1])
# sys.exit()
print("====== done step 1 =======")
npts = 30
pstraini = []
new_xx = result.x[:, nby0:nby0 + nby1]
new_yy = result.y[:, nby0:nby0 + nby1]
new_result = result.vx[:, nby0:nby0 + nby1]
for (xpt, ypt) in zip(new_xx.flatten(), new_yy.flatten()):
dist = np.sqrt((x - xpt)**2 + (y - ypt)**2)
order = dist.argsort()
f1 = interpolate.SmoothBivariateSpline(x[order[:npts]],
y[order[:npts]],
pstrain[order[:npts]],
kx=3,
ky=3)
pstrainres = f1(xpt, ypt)
pstraini.append(pstrainres[0][0])
print("====== done step 2 =======")
pstraini = np.reshape(np.array(pstraini), np.shape(new_result))
print(np.shape(pstraini))
sys.exit()
return pstraini
def panei_mr(targetTemp, baseDir):
'''given a target temp, returns a function giving
radius as a function of mass.
function is derived from cubic interpolation of Panei models'''
assert np.all((targetTemp>=5000*units.K)&(targetTemp<45000*units.K)), \
"Model invalid at temps less than 4000 or greater than 45,000 K"
# read panei model grid in
teffs,masses,radii = \
read_panei_file(os.path.join(baseDir,'Panei/12C-MR-H-He/12C-MR-H-He.dat'))
# this function interpolates to give radius as function of temp, mass
func = interp.SmoothBivariateSpline(teffs, masses, radii)
# this function specifies the target temp to give just radius as fn of mass
f2 = lambda x: func(targetTemp, x)[0]
return f2
def test_interpolation(mesh):
from scipy import interpolate
x, y = mesh.x, mesh.y
Z = np.exp(-x**2 - y**2)
# We ensure interpolation points are within convex hull
xi = np.random.uniform(0.1, 0.9, 10)
yi = np.random.uniform(0.1, 0.9, 10)
# Stripy
zn1 = mesh.interpolate_nearest(xi, yi, Z)
zl1, err = mesh.interpolate_linear(xi, yi, Z)
zc1, err = mesh.interpolate_cubic(xi, yi, Z)
# cKDTree
tree = interpolate.NearestNDInterpolator((x, y), Z)
zn2 = tree(xi, yi)
# Qhull
tri = interpolate.LinearNDInterpolator((x, y), Z, 0.0)
zl2 = tri((xi, yi))
# Clough Tocher
cti = interpolate.CloughTocher2DInterpolator(np.column_stack([x,y]),\
Z, tol=1e-10, maxiter=20)
zc2 = cti((xi, yi))
zc2[np.isnan(zc2)] = 0.0
# Spline
spl = interpolate.SmoothBivariateSpline(x, y, Z)
zc3 = spl.ev(xi, yi)
# Radial basis function
rbf = interpolate.Rbf(x, y, Z)
zc4 = rbf(xi, yi)
print("squared residual in interpolation\n \
- nearest neighbour = {}\n \
- linear = {}\n \
- cubic (clough-tocher) = {}\n \
- cubic (spline) = {}\n \
- cubic (rbf) = {}" .format(((zn1 - zn2)**2).max(), \
((zl1 - zl2)**2).max(), \
((zc1 - zc2)**2).max(), \
((zc1 - zc3)**2).max(), \
((zc1 - zc4)**2).max(),) )
def get_simple_interpolation_fn(self, axis=MEDIAL):
"""
Returns a simple spline interpolation callable fitted across the whole field.
:param axis: Pass constant SAGGITAL, MERIDIONAL, or MEDIAL
:return: Scipy callable which accepts x and y positions and provides interpolated value.
"""
lst = []
for point in self.get_subset(axis):
lst.append((point.x, point.y, point.mtf50))
x_lst, y_lst, z_lst = zip(*lst)
fn = interpolate.SmoothBivariateSpline(x_lst,
y_lst,
z_lst,
kx=2,
ky=2,
s=float("inf"))
return fn
def test_derivative(mesh):
from scipy import interpolate
# Create a field to test derivatives
x, y = mesh.x, mesh.y
Z = np.exp(-x**2 - y**2)
Zx = -2 * x * Z
Zy = -2 * y * Z
gradZ = np.hypot(Zx, Zy)
# Stripy
t = clock()
Zx1, Zy1 = mesh.gradient(Z, nit=10, tol=1e-10)
t1 = clock() - t
gradZ1 = np.hypot(Zx1, Zy1)
# Spline
spl = interpolate.SmoothBivariateSpline(x, y, Z)
t = clock()
Zx2 = spl.ev(x, y, dx=1)
Zy2 = spl.ev(x, y, dy=1)
t2 = clock() - t
gradZ2 = np.hypot(Zx2, Zy2)
# Clough Tocher
# This one is most similar to what is used in stripy
t = clock()
cti = interpolate.CloughTocher2DInterpolator(np.column_stack([x,y]),\
Z, tol=1e-10, maxiter=20)
t3 = clock() - t
Zx3 = cti.grad[:, :, 0].ravel()
Zy3 = cti.grad[:, :, 1].ravel()
gradZ3 = np.hypot(Zx3, Zy3)
res1 = ((gradZ1 - gradZ)**2).max()
res2 = ((gradZ2 - gradZ)**2).max()
res3 = ((gradZ3 - gradZ)**2).max()
print("squared error in first derivative\n \
- stripy = {} took {}s\n \
- spline = {} took {}s\n \
- cloughtocher = {} took {}s".format(res1, t1, res2, t2, res3, t3))
def fit_spline( x, y, x_ref, knot_x=None, knot_y=None, num_knots=5, smooth=False, smooth_fac=None, weights=None, order=3):
'''
performs spline fit of the form dx = f(x,y)
knot_x/knot_y are 1-d arrays that are the x and y coordinates of knot locations
'''
if knot_x == None:
knot_x = np.linspace(np.min(x), np.max(x), num=num_knots)
if knot_y == None:
knot_y = np.linspace(np.min(y), np.max(y), num=num_knots)
if not smooth:
if smooth_fac == None:
spline = interpolate.LSQBivariateSpline(x, y, x_ref, knot_x, knot_y, kx=order, ky=order, w=weights)
x_new = spline.ev(x, y)
else:
spline = interpolate.LSQBivariateSpline(x, y, x_ref, knot_x, knot_y, kx=order, ky=order,w=weights, s=smooth_fac)
x_new = spline.ev(x, y)
else:
spline = interpolate.SmoothBivariateSpline(x, y, x_ref, w=weights)
x_new = spline.ev(x, y)
return x_new, spline
def interp_2d_values_from_profiles(self):
"""Interpolate values in 2D from profiles"""
ux = np.array([], dtype=np.float)
vy = np.array([], dtype=np.float)
new_xt = self.points['xt']
new_xl = self.points['xl'] + self.points['zone']
new_values = np.zeros((self.nb_var, len(self.points)))
for i, profile in enumerate(self.section_seq):
first_xt = profile.get_limit_by_idx(0)['Xt_profil']
last_xt = profile.get_limit_by_idx(-1)['Xt_profil']
xt = (profile.coord.array['Xt'] - first_xt) / (last_xt - first_xt)
ux = np.concatenate((ux, xt))
vy = np.concatenate((vy, np.array([i] * profile.nb_points)))
for j, var in enumerate(self.var_names()):
z = np.array([], dtype=np.float)
for profile in self.section_seq:
z = np.concatenate((z, profile.coord.values[var]))
if self.interp_values == 'BIVARIATE_SPLINE':
interp_bivariate_spline = interpolate.SmoothBivariateSpline(
ux, vy, z, kx=3, ky=3)
for k, (u, v) in enumerate(zip(new_xt, new_xl)):
new_values[j, k] = interp_bivariate_spline(u, v)[0][0]
else:
if self.interp_values == 'BILINEAR':
method = 'linear'
elif self.interp_values == 'BICUBIC':
method = 'cubic'
else:
raise NotImplementedError
new_values[j, :] = interpolate.griddata((ux, vy),
z, (new_xt, new_xl),
method=method)
return new_values
def CubicSplineInterpolation(data = None, mask = None,filePath= ""):
# griddata()
results = np.zeros(data.shape)
for i in range(data.shape[0]):
if i % 100 == 0:
print(i)
temp = np.zeros((int(np.sum(mask[i,:,:,0].flatten())),3))
index = 0
for j in range(data.shape[1]):
for k in range(data.shape[2]):
if mask[i,j,k,0] == 1:
temp[index,0] = k
temp[index,1] = j
temp[index,2] = data[i,j,k,0]
index = index + 1
x = np.linspace(0, 40, 41)
y = np.linspace(0, 40, 41)
# print(temp[:,2].max())
# x, y = np.meshgrid(x, y)#20*20的网格数据
# z = interpolate.griddata(temp[:,0:2], temp[:,2], (x, y), method='cubic')
# func = interpolate.interp2d(temp[:,0], temp[:,1], temp[:,2], kind='cubic')
w = np.ones((len(temp[:,0]),1))
zum = np.sum(temp[:,2]**2)
smooth = 0.01
print(zum)
func = interpolate.SmoothBivariateSpline(temp[:,1], temp[:,0], temp[:,2], w=w, s = 0)
z = func(x, y)
results[i,:,:,0] = z
# plt.subplot(121)
# print(z.shape)
# plt.imshow(z)
# plt.subplot(122)
# plt.imshow(data[i,:,:,0])
# plt.show()
np.save(filePath, results)
return results
yoff)
continue
else:
smean[missmean] = scoarse[missmean]
smasknotmiss = np.logical_and(np.logical_not(miss), smask)
if (np.sum(smasknotmiss) == 0):
smean = scoarse.copy()
sanddiffs = ldat[smasknotmiss] - smean[smasknotmiss]
## if (np.sum(smasknotmiss) > 5):
## fill = np.interp(np.nonzero(np.logical_not(smasknotmiss)), np.nonzero(smasknotmiss), sanddiffs)
## else:
## continue
goodpnt = np.nonzero(np.logical_not(smasknotmiss))
fillpnt = np.nonzero(smasknotmiss)
if (np.sum(smasknotmiss) > 0):
interpfunc = interpolate.SmoothBivariateSpline(
goodpnt[0], goodpnt[1], sanddiffs)
fill = interpfunc.ev(fillpnt[0], fillpnt[1])
adj = np.zeros((256, 256))
adj[goodpnt] = sanddiffs
adj[fillpnt] = fill
else:
adj = np.zeros((256, 256))
## adj[np.logical_not(smasknotmiss)] = fill
## smalldiff = np.logical_and(np.less(ldat, ldat + (sandsdev*2)), np.greater(ldat, ldat - (sandsdev*2)))
## goodgood = np.logical_and(smask, smalldiff)
ldatadj = ldat - adj
# add one for each deviation above baseline
n_dev[ldatadj > baseline] += 1
## ldatadj[np.logical_not(smalldiff)] = 0
# save the maximum deviation
def compute_Delta(redmap_cluster_file, redmap_member_file, output_file,
num_lam_bins, num_z_bins):
#Compute Delta given redmap catalog and redmapmember catalog
print "Reading redmapper catalog..."
# Data catalog
try:
hdu = pf.open(redmap_cluster_file)
memmatchid_clusters = hdu[1].data.field('mem_match_id')
lam_clusters = hdu[1].data.field('lambda_chisq')
z_clusters = hdu[1].data.field('z_lambda')
ra_clusters = hdu[1].data.field('ra')
dec_clusters = hdu[1].data.field('dec')
hdu.close()
except:
print "Invalid redmapper cluster file!"
num_clusters = len(lam_clusters)
print "Reading redmapper member catalog..."
# Read in member catalog
try:
hdu = pf.open(redmap_member_file)
memmatchid_members = hdu[1].data.field('mem_match_id')
pfree_members = hdu[1].data.field('pfree')
p_members = hdu[1].data.field('p')
r_members = hdu[1].data.field('r')
theta_i_members = hdu[1].data.field('theta_i')
theta_r_members = hdu[1].data.field('theta_r')
hdu.close()
except:
print "Invalid redmapper member file!"
num_members = len(r_members)
#Compute membership probabilities
memprob_members = pfree_members * p_members * theta_i_members * theta_r_members
print "Calculating meanr..."
# Calculate meanr for every cluster
# This stores <R_mem> for every cluster
meanr_clusters = np.zeros(num_clusters) - 1.
memmatchid_clusters_output = np.zeros(num_clusters) - 1.
counter = 0
prev_index = 0
for mi in xrange(1, num_members):
if (mi % 1000000 == 0):
print "member = ", mi + 1, " out of ", num_members + 1
if (memmatchid_members[mi] != memmatchid_members[prev_index]):
meanr_clusters[counter] = np.sum(
memprob_members[prev_index:mi] *
r_members[prev_index:mi]) / np.sum(
memprob_members[prev_index:mi])
memmatchid_clusters_output[counter] = memmatchid_members[
prev_index]
counter += 1
prev_index = mi
# Last cluster done separately
meanr_clusters[counter] = np.sum(
memprob_members[prev_index:] * r_members[prev_index:]) / np.sum(
memprob_members[prev_index:])
memmatchid_clusters_output[counter] = memmatchid_members[prev_index]
# Test for mismatch
bad = np.where(memmatchid_clusters_output != memmatchid_clusters)[0]
if (len(bad) > 0):
raise Exception("Mismatching indices!")
# Compute average meanr in bins of richness and redshift
lam_bins_meanr = np.exp(
np.linspace(np.log(0.999 * np.min(lam_clusters)),
np.log(1.001 * np.max(lam_clusters)),
num=num_lam_bins + 1))
z_bins_meanr = np.linspace(0.999 * np.min(z_clusters),
1.001 * np.max(z_clusters),
num=num_z_bins + 1)
meanr_mat = np.zeros((num_lam_bins, num_z_bins))
print "Computing meanr across grid..."
for li in xrange(0, num_lam_bins):
for zi in xrange(0, num_z_bins):
in_bin = np.where((lam_clusters > lam_bins_meanr[li])
& (lam_clusters <= lam_bins_meanr[li + 1])
& (z_clusters > z_bins_meanr[zi])
& (z_clusters <= z_bins_meanr[zi + 1]))[0]
#Only compute mean if there are clusters in bin
if (len(in_bin) > 0):
meanr_mat[li, zi] = np.mean(meanr_clusters[in_bin])
print "Computing average value of meanr as a function of richness and z..."
# Get average value of meanr for clusters of that richness and redshift, evaluated at richness and redshift of all clusters
# Use simple interpolation of grid
avg_meanr_clusters_simple = interp_avg_meanr(lam_clusters, z_clusters,
lam_bins_meanr, z_bins_meanr,
meanr_mat)
# Get avg meanr using ungridded spline fit
spline = interpolate.SmoothBivariateSpline(lam_clusters, z_clusters,
meanr_clusters)
avg_meanr_clusters_ungridspline = spline.ev(lam_clusters, z_clusters)
# Get avg meanr using gridded spline fit
meanr_mat = np.nan_to_num(meanr_mat)
rbs = interpolate.RectBivariateSpline(
np.exp(0.5 *
(np.log(lam_bins_meanr[1:]) + np.log(lam_bins_meanr[:-1]))),
0.5 * (z_bins_meanr[1:] + z_bins_meanr[:-1]),
meanr_mat,
ky=2)
avg_meanr_clusters_gridspline = rbs.ev(lam_clusters, z_clusters)
#Compute Delta for all clusters
avg_meanr_clusters = np.copy(avg_meanr_clusters_gridspline)
Delta_clusters = (meanr_clusters - avg_meanr_clusters) / avg_meanr_clusters
print "Outputting to fits file..."
#Create columns for fits file
ra_col = pf.Column(name='RA', format='E', array=ra_clusters)
dec_col = pf.Column(name='DEC', format='E', array=dec_clusters)
lam_col = pf.Column(name='lambda_chisq', format='E', array=lam_clusters)
z_col = pf.Column(name='z_lambda', format='E', array=z_clusters)
meanr_col = pf.Column(name='R_mem', format='E', array=meanr_clusters)
avg_meanr_col = pf.Column(name='mean_R_mem',
format='E',
array=avg_meanr_clusters)
Delta_col = pf.Column(name='Delta', format='E', array=Delta_clusters)
#All columns
cols = pf.ColDefs(
[ra_col, dec_col, lam_col, z_col, meanr_col, avg_meanr_col, Delta_col])
#Write to fits
tbhdu = pf.TableHDU.from_columns(cols)
try:
tbhdu.writeto(output_file)
except:
print "Unable to write to output file!"
raise
def make_contour_plot_multrow(Xs,
Ys,
Cs,
elevation=30,
azimuthal=30,
alpha=0.8,
xlabel='',
ylabel='',
zlabel='',
clabel='',
title='',
titles=[],
bounds=None,
colors=[],
equal_mass=1,
colorbartype='simple',
label=None,
logC=True,
cmin=0.8,
cmax=1.,
savefig='plots/plot.png'):
# {{{
if not len(Xs) == len(Ys) == len(Cs):
raise IOError("Length of lists to be plotted not the same")
# Known failure mode: when all arrays in Xrows are of the same length
nrows, ncols = np.shape(Xs)
print("No of rows = %d, columns = %d" % (nrows, ncols))
#
gmean = (5.**0.5 - 1) * 0.5
if logC:
bounds = np.log10(bounds)
print("Using logscale on Z", file=sys.stderr)
for ridx, C in enumerate(Cs):
for cidx, R in enumerate(C):
Cs[ridx][cidx] = np.log10(R)
if 'FF' in clabel:
cmin = -2.3
else:
Cs = np.array(Cs)
print("Shape of C = ", np.shape(Cs), "dtype of C = ", type(Cs))
cmin = np.inf
for tmpr in Cs:
for tmpc in tmpr:
print(np.shape(tmpc), np.shape(tmpr))
if cmin > np.min(tmpc):
cmin = np.min(tmpc)
print("Min = ", cmin)
cmin = np.round(cmin, decimals=3)
#clabel = clabel + ' (Log)'
else:
print("NOT using logscale on Z", file=sys.stderr)
if 'FF' in clabel:
cmin = 10**-2.3
else:
Cs = np.array(Cs)
print("Shape of C = ", np.shape(Cs), "dtype of C = ", type(Cs))
cmin = np.inf
for tmpr in Cs:
for tmpc in tmpr:
cmin = min(cmin, np.min(tmpc))
print("Min = ", cmin)
cmin = np.round(cmin, decimals=3)
cmax = -np.inf
for tmpr in Cs:
for tmpc in tmpr:
cmax = max(cmax, np.max(tmpc))
cmax = np.round(cmax, decimals=3)
#
if bounds != None and (type(bounds) == np.ndarray or type(bounds) == list):
print("bounds before insert : ", bounds)
bounds = insert_min_max_into_array(bounds, cmin, cmax)
print("bounds after insert : ", bounds)
#
# Insert default values of bounds
if bounds is None and logC:
if 'FF' in clabel or 'mathcal{M}' in clabel:
bounds = np.log10([0.0001, 0.005, 0.01, 0.02, 0.03, 0.05, 0.1, 1])
elif '\mathcal{M}_c' in clabel:
print("CHIRP MASS PLOT")
bounds = np.log10([0.005, 0.01, 0.02, 0.03, 0.05, 0.1, 0.2])
elif '\Delta' in clabel:
bounds = np.log10([0.005, 0.01, 0.03, 0.05, 0.1, 0.2, 0.5, 1.])
else:
bounds = np.linspace(cmin, cmax, 10)
#bounds = np.append( bounds - 0.1, 0 )
elif bounds is None:
raise IOError("Non-log default colorbar bounds not supported")
for tid, t in enumerate(title):
t[tid].replace('_', '-')
fig = plt.figure(int(1e7 * np.random.random()),
figsize=(4 * ncols, 4 * nrows))
# cmap = plt.cm.RdYlGn_r#gist_heat_r##winter#gnuplot#PiYG_r#RdBu_r#jet#rainbow#RdBu_r#Spectral_r
# cmap = plt.cm.PiYG_r#rainbow#RdBu_r#RdYlGn_r
#cmap = plt.get_cmap('jet_r', 20)
cmap = plt.cm.OrRd
#cmaplist = [cmap(i) for i in range(cmap.N)]
#cmap = cmap.from_list('Custom map', cmaplist, cmap.N)
# cmap.set_under('gray')
print("bounds = ", bounds)
if type(bounds) == np.ndarray or type(bounds) == list:
norm = mp.colors.BoundaryNorm(bounds, cmap.N)
else:
tmp_bounds = np.linspace(cmin, cmax, 10)
norm = mp.colors.BoundaryNorm(tmp_bounds, cmap.N)
#
# Begin plotting loop
nplot = 0
allaxes = []
for rowid in range(nrows):
for colid in range(ncols):
nplot += 1
ax = fig.add_subplot(nrows, ncols, nplot)
allaxes.append(ax)
#
X, Y, C = Xs[rowid][colid], Ys[rowid][colid], Cs[rowid][colid]
print(np.shape(X), np.shape(Y), np.shape(C))
# Add points in the bottom plane marking spins
if equal_mass == colid + 1:
tmpX, tmpY = X, Y
X = np.append(X, tmpY)
Y = np.append(Y, tmpX)
C = np.append(C, C)
#
Xrange = np.linspace(min(X), max(X), 2000)
Yrange = np.linspace(min(Y), max(Y), 2000)
#Xrange = np.linspace( -1, 1, 100 )
#Yrange = np.linspace( -1, 1, 100)
Xmap, Ymap = np.meshgrid(Xrange, Yrange)
print(np.shape(X), np.shape(Y), np.shape(C))
colormap = plt.mlab.griddata(X, Y, C, Xmap, Ymap, interp='linear')
#
import scipy.interpolate as si
rbfi = si.SmoothBivariateSpline(X, Y, C, kx=4, ky=4)
#colormap = rbfi(Xrange, Yrange)
#
# New interpolation scheme
#
#xyzData = np.append( np.append( [X], [Y], axis=0 ), [C], axis=0 )
#xyzData = scipy.ndimage.zoom(xyzData, 3)
#Xmap = xyzData[:,0]
#Ymap = xyzData[:,1]
#colormap = xyzData[:,2]
print("Shape pof Xmap, Ymap, colormap = ", np.shape(Xmap),
np.shape(Ymap), np.shape(colormap))
#Xmap = scipy.ndimage.zoom(Xmap, 3)
#Ymap = scipy.ndimage.zoom(Ymap, 3)
#colormap = scipy.ndimage.zoom(colormap, 3)
if len(colors) == (len(bounds) - 1):
CS = ax.contourf(Xmap, Ymap, colormap,
levels=bounds,
colors=colors,
alpha=0.75,\
# cmap=plt.cm.spectral,\
linestyles='dashed')
'''CS = ax.tricontourf(X,Y,C,\
levels=bounds,\
colors=colors,\
alpha=0.9)'''
'''CS1 = ax.scatter(X, Y, c='k', s=5)#, \
#reduce_C_function=np.max)'''
else:
ax.contourf(Xmap,
Ymap,
colormap,
bounds,
cmap=cmap,
linestyles='dashed')
ax.set_xlabel(xlabel)
ax.set_ylabel(ylabel)
ax.set_xlim([-1, 1])
ax.set_ylim([-1, 1])
# ax.zaxis.set_rotate_label(False)
#ax.set_zlabel(zlabel, rotation=90)
print("Len(titles) = %d, NCOLS = %d" % (len(titles), ncols))
if len(titles) == ncols:
ax.set_title(titles[colid], verticalalignment='bottom')
elif colid == ncols / 2:
ax.set_title(title[rowid] + '\n $q=%d$' % (colid + 1),
verticalalignment='bottom')
else:
ax.set_title('$q=%d$' % (colid + 1),
verticalalignment='bottom')
ax.grid()
#
if colorbartype == 'simple':
ax2 = fig.add_axes([0.92, 0.1, 0.01, 0.7])
cb = plt.colorbar(CS, cax=ax2, orientation=u'vertical', format='%.3f')
cb.set_label(clabel)
else:
ax2 = fig.add_axes([0.92, 0.1, 0.01, 0.7])
#ax2 = fig.add_axes([0.2, 0.05, 0.6, 0.02])
# Make the colorbar
if type(bounds) == np.ndarray or type(bounds) == list:
cb = mp.colorbar.ColorbarBase(ax2,
cmap=cmap,
norm=norm,
spacing='uniform',
format='%.2f',
orientation=u'vertical',
ticks=bounds,
boundaries=bounds)
else:
# How does this colorbar know what colors to span??
cb = mp.colorbar.ColorbarBase(ax2,
cmap=cmap,
norm=norm,
spacing='uniform',
format='%.2f',
orientation=u'vertical',
ticks=tmp_bounds)
# Add tick labels
if logC and (type(bounds) == np.ndarray or type(bounds) == list):
cb.set_ticklabels(np.round(10**bounds, decimals=4))
elif type(bounds) == np.ndarray or type(bounds) == list:
cb.set_ticklabels(np.round(bounds, decimals=4))
# cb = fig.colorbar(scat, shrink=0.5, aspect=30, spacing='proportional',\
# ticks=[0,-0.5,-1,-1.3,-1.6,-1.9,-2.1,-2.4,-2.7,-3])
#ax2.set_title(clabel, loc='left')
# ,labelpad=-0.3,y=1.1,x=-0.5)
cb.set_label(clabel,
verticalalignment='top',
horizontalalignment='center')
#if max(C) < cmax and min(C) > cmin: cb.set_clim([cmin,cmax])
fig.tight_layout(rect=(0, 0, 0.93, 1))
if '.png' in savefig:
savefig = savefig.split('.png')[0] + '_q123.png'
elif '.pdf' in savefig:
savefig = savefig.split('.pdf')[0] + '_q123.pdf'
fig.savefig(savefig)
return
def estimate_alpha_and_beta(input_filepath, quantity_to_compute, boundary,
radius, grid_size, value_change, method_angle,
method_curv, region_of_interest, region_points):
"""
Imports a matrix of parameter values corresponding
to a (alpha,beta) point and perform spline interpolation
to make a parameter surface.
Parameter surface is used to find intersection with
initial parameter value plus / minus one standard deviation.
Three criteria to find suggested alpha-beta value from intersection.
Args:
input_filepath (str): Surface model filename and location where data is stored.
quantity_to_compute(str): Parameter name.
boundary (list): Boundary of searching grid.
radius (float): Minimum radius of circle to search outside of.
grid_size (int): Size of searching grid ( grid_size x grid_size matrix)
value_change (float): Desired change in curvature / bend angle to achieve
method_angle (str): Method for computing angle.
method_curv (str): Method for computing curvature.
region_of_interest (str): Method for setting the region of interest ['manual' | 'commandline' | 'landmarking']
region_points (list): If region_of_interest is 'commandline', this a flatten list of the start and endpoint
"""
# Get grid values
base_path = get_path_names(input_filepath)
# Get region points
# Compute discrete values for quantity
if type(boundary[-1]) is str:
boundary = np.asarray(boundary, dtype=float)
data = compute_quantities(input_filepath,
boundary,
quantity_to_compute,
method_curv,
method_angle,
region_of_interest,
region_points,
n=grid_size,
projection=False)
# Get grid boundary
amin, amax, bmin, bmax = float(boundary[0]), float(boundary[1]), float(
boundary[2]), float(boundary[3])
# Set standard deviations used to find intersection
# Defined SD planes for curvature
# Tolerance added for adjusting SD
# if there are no intersections found
def value_plus(tolerance=0.0):
return initial_value + value_change - tolerance
def value_minus(tolerance=0.0):
return initial_value - value_change + tolerance
n = len(data)
alpha = np.linspace(amin, amax, n)
beta = np.linspace(bmin, bmax, n)
alpha_long = np.linspace(amin, amax, 300)
beta_long = np.linspace(bmin, bmax, 300)
xx, yy = np.meshgrid(alpha, beta)
points = np.zeros((n, 2))
for i in range(len(xx)):
points[i] = [alpha[i], beta[i]]
# Spline interpolation
f = interpolate.SmoothBivariateSpline(xx.ravel(), yy.ravel(), data.ravel())
initial_value = f(0, 0)
methods = [value_plus, value_minus]
# Find intersecting points
# Reduces SD if no points are found
for plane in methods:
zeros = alpha_beta_intersection(plane, f, alpha_long, beta_long)
if len(zeros) == 0:
empty = True
# Leeway tolerance for matching quantity on interpolated surface
tol = 0.005 if quantity_to_compute == "curvature" else 0.1
max_iter = 50
iterations = 0
print("-- Found no points..Adjusting SD")
while empty and iterations < max_iter:
print("-- Iterations: %i" % (iterations + 1))
zeros = alpha_beta_intersection(plane, f, alpha_long,
beta_long, tol)
if len(zeros) > 0:
empty = False
iterations += 1
if quantity_to_compute == "curvature":
tol += 0.001
elif quantity_to_compute == "angle":
tol += 0.2
# Check points and apply criteria
# to find suggested values for alpha and beta
if len(zeros) > 0:
points = []
for p in zeros:
if (plane.__name__ == "value_plus" and quantity_to_compute == "curvature") \
or (plane.__name__ == "value_minus" and quantity_to_compute == "angle"):
if p[1] < 0:
points.append(p)
else:
if p[1] > 0:
points.append(p)
suggested_point = points[0]
dist = 1e9
for p in points[1:]:
dist_tmp = la.norm(np.array(p))
if radius < dist_tmp < dist:
dist = dist_tmp
suggested_point = p
# Write points to file
write_alpha_beta_point(base_path, suggested_point, plane.__name__,
quantity_to_compute)
def evaluate_quote(self, longitudes, latitudes, elevations, x_point,
y_point):
dists = np.sqrt(
np.sum((np.stack(
(longitudes, latitudes), axis=-1) - [x_point, y_point])**2, 1))
if is_scipy:
print 'use scipy to interpolate'
#tck = interpolate.splrep(x, y, s=0)
#xnew = np.linspace(np.min(x), np.max(x), 200)
#ynew = interpolate.splev(xnew, tck, der=0)
#if 1:
nearest_longitudes = longitudes[(dists <
self.interpolation_radius)]
nearest_latitudes = latitudes[(dists < self.interpolation_radius)]
nearest_elevations = elevations[(dists <
self.interpolation_radius)]
## nearest_longitudes = longitudes[(longitudes < x_point + self.interpolation_radius)&(longitudes > x_point - self.interpolation_radius)&(latitudes < y_point + self.interpolation_radius)&(latitudes > y_point - self.interpolation_radius)]
## nearest_latitudes = latitudes[(longitudes < x_point + self.interpolation_radius)&(longitudes > x_point - self.interpolation_radius)&(latitudes < y_point + self.interpolation_radius)&(latitudes > y_point - self.interpolation_radius)]
## nearest_elevations = elevations[(longitudes < x_point + self.interpolation_radius)&(longitudes > x_point - self.interpolation_radius)&(latitudes < y_point + self.interpolation_radius)&(latitudes > y_point - self.interpolation_radius)]
print[x_point, y_point
], nearest_longitudes, nearest_latitudes, nearest_elevations
if len(nearest_longitudes) > 15:
f_inter = interpolate.SmoothBivariateSpline(
nearest_longitudes,
nearest_latitudes,
nearest_elevations,
) #kind = self.method_interp )
## #############################################
## xnew = np.linspace(x_point-self.interpolation_radius/2, x_point+self.interpolation_radius/2,200)
## ynew = np.linspace(y_point-self.interpolation_radius/2, y_point+self.interpolation_radius/2,200)
## X, Y = np.meshgrid(xnew, ynew)
## Z = f_inter(xnew, ynew)
##
##
## fig = plt.figure()
## ax = fig.gca(projection='3d')
##
##
## # Plot the surface.
## surf = ax.plot_surface(X, Y, Z, cmap=cmp.coolwarm,
## linewidth=0, antialiased=False)
##
## fig.colorbar(surf, shrink=0.5, aspect=5)
## plt.savefig('/home/cristian/scenarios_cri/Elevation/interpolation_%d'%(xnew[0]))
## plt.show()
#############################################
quote = f_inter(x_point, y_point)
else:
quote = elevations[np.argmin(dists)]
print 'nearest quote'
else:
nearest_quotes = elevations[(dists < 100)]
nearest_dists = dists[(dists < 100)]
numerator = 0.0
denominator = 0.0
for near_quote, near_dist in zip(nearest_quotes, nearest_dists):
numerator += near_quote / (10**(near_dist / 10))
denominator += 1 / (10**(near_dist / 10))
## print numerator, denominator
if denominator != 0:
quote = numerator / denominator
else:
quote = elevations[np.argmin(dists)]
print 'nearest quote'
return quote
def interp_hazard_curve(sitelon, sitelat, hazcurvefile):
from openquake.nrmllib.hazard.parsers import HazardCurveXMLParser
from numpy import array, where
from scipy import interpolate
import warnings
warnings.filterwarnings("ignore")
hcm = HazardCurveXMLParser(hazcurvefile).parse()
#curves = hcc._set_curves_matrix(hcm)
#extract lat/lons from POES
curlat = []
curlon = []
curves = []
for loc, poes in hcm:
curlon.append(loc.x)
curlat.append(loc.y)
curves.append((poes))
curlon = array(curlon)
curlat = array(curlat)
curves = array(curves)
# check to see if site within model
if sitelat >= min(curlat) and sitelat <= max(curlat) \
and sitelon >= min(curlon) and sitelon <= max(curlon):
# find indexes about site
index1 = where((curlon >= sitelon-1.) & (curlon <= sitelon+1.) \
& (curlat >= sitelat-1.) & (curlat <= sitelat+1.))[0]
lonstrip = array(curlon)
latstrip = array(curlat)
curstrip = array(curves)
# build array of poes for each return period
interphazcurve = []
for i in range(0, len(curstrip[0])):
poearray = []
for c in curstrip:
poearray.append(c[i])
# standard 2D interpolation does not work!
'''
# do 2D interpolation across log values
interpfunc = interpolate.interp2d(lonstrip, latstrip, \
log(array(poearray)), kind='linear')
interphazcurve.append(exp(interpfunc(sitelon, sitelat)[0]))
'''
# use SmoothBivariateSpline instead
'''
# smooth ln hazard - changed to linear b/c cannot interp log(0)
interpfunc = interpolate.SmoothBivariateSpline(lonstrip, latstrip, \
log(array(poearray)))
# evaluate SmoothBivariateSpline
interphazcurve.append(exp(interpfunc.ev(sitelon, sitelat)))
'''
# smooth hazard
interpfunc = interpolate.SmoothBivariateSpline(lonstrip, latstrip, \
array(poearray))
# evaluate SmoothBivariateSpline
interphazcurve.append(interpfunc.ev(sitelon, sitelat))
return array(interphazcurve), hcm.metadata, curstrip
def intp_comparison(isotope,option,size,Grid_Setting,Test_type,space_factor,
op_mode,X,Y,sorting_index,Kernel_type,Xsobol,Ysobol):
"""
Comparison between Cubic Splines and GP models.
inputs:
*isotope: nuclide to use for comparison
*option:
-Grid: for comparisons on models trained on a Grid
-Random: for comparisons on models trained on a random selection
of the datasets.
-Sobol: for comparisons on models trained on a Sobol sequence.
*size: The number of samples to consider. Only considered when selecting
the option 'Sobol' or 'Random'.
*Grid_Setting: Selects the option on the design generator (Even, Odd or
Checkerboard).
*Test_type: Selects the option on the design generator(All, inner or
borders).
*space_factor: Sets the spacing on the design generator.
*op_mode: Sets the operation mode of the design generator (normal,
testing).
* X,Y, Xsobol, Ysobol: Training / Testing sets
* sorting_index: Only used for X and Y on a grid. Reorder Ys points to
correctly match the points of the sampling grid.
* Kernel_type: Used to select the kernel type used (Kernels-Mass,
Kernels-Adens. etc)
outputs:
* Values_Cubic,Values_GPR: Contain diagnostics such as MAE, RMSE, MSE,
Rsquared, plus mean predictive variance and
fractions of predicted points within
predictive variance for GPR.
"""
Temperature = np.array(list(set(X[:,0])))
Burnup = np.array(list(set(X[:,1])))
Burnup = np.sort(Burnup)
Ys = np.array(Y[isotope])[sorting_index]
Ys = Ys.reshape((len(Temperature),len(Burnup))).T
Ydata = np.array(Y[isotope])
if option == 'Grid':
# =========================================================================
# Cubic
# =========================================================================
Tmesh, Bmesh = np.meshgrid(Temperature,Burnup)
BtrainCubic,BtestCubic = gen_mask(
Grid_Setting,Bmesh,Test_type,space_factor,op_mode)
TtrainCubic,TtestCubic = gen_mask(
Grid_Setting,Tmesh,Test_type,space_factor,op_mode)
YtrainCubic,YtestCubic = gen_mask(
Grid_Setting,Ys,Test_type,space_factor,op_mode)
# =========================================================================
# GPR
# =========================================================================
XtrainGPR = np.vstack((TtrainCubic,BtrainCubic)).T
YtrainGPR = YtrainCubic
XtestGPR = np.vstack((TtestCubic,BtestCubic)).T
YtestGPR = YtestCubic
elif option == 'random':
Indexes = np.arange(len(X))
IdxTrain = np.random.choice(Indexes,size = size)
IdxTest = np.array([idx for idx in Indexes if idx not in IdxTrain])
XtrainGPR = X[IdxTrain]
XtestGPR = X[IdxTest]
BtrainCubic = XtrainGPR[:,0]
TtrainCubic = XtrainGPR[:,1]
BtestCubic = XtestGPR[:,0]
TtestCubic = XtestGPR[:,1]
YtrainGPR = Ydata[IdxTrain]
YtestGPR = Ydata[IdxTest]
YtrainCubic = YtrainGPR
YtestCubic = YtestGPR
elif option == 'Sobol':
XtrainGPR = Xsobol[:size]
YtrainGPR = Ysobol[:size]
XtestGPR = X
YtestGPR = Y[isotope]
BtrainCubic = XtrainGPR[:,1]
TtrainCubic = XtrainGPR[:,0]
BtestCubic = XtestGPR[:,1]
TtestCubic = XtestGPR[:,0]
YtrainCubic = YtrainGPR
YtestCubic = YtestGPR
# =========================================================================
# Interpolators
# =========================================================================
CubicInt = intp.SmoothBivariateSpline(BtrainCubic,TtrainCubic,YtrainCubic)
Yintp_Cubic = CubicInt.ev(BtestCubic,TtestCubic)
Cubic_Errors = get_Error(
Yintp_Cubic,YtestCubic)
# =========================================================================
# GPR:
# =========================================================================
try:
# =====================================================================
# Load Kernel Params:
# =====================================================================
Kernel = np.load('Path/'+Kernel_type+\
'/{}/{}.npy'.format(option,isotope),
allow_pickle=True).item()
Params = Kernel['Params']
Lambda_Inv = Kernel['LAMBDA']
# =====================================================================
# Precalculate alpha_ and K_inv:
# =====================================================================
alpha_,K_inv = alpha_calculator(Kernel['Params'],XtrainGPR,YtrainGPR)
# =====================================================================
# Get predictions:
# =====================================================================
GPR = [GPR_MODEL_w_Std(Params,Lambda_Inv,XtrainGPR,YtrainGPR,alpha_,
K_inv,x) for x in XtestGPR]
GPR_pred = np.array(GPR)[:,0]
GPR_std = np.array(GPR)[:,1]
GPR_Errors = get_Error(GPR_pred,YtestGPR)
except FileNotFoundError:
return 404
Mean_GPR_std = np.mean(GPR_std)
Max_GPR_std = np.max(GPR_std)
RsquareGPR = Rsquare(GPR_pred,YtestGPR)
RsquareCubic = Rsquare(Yintp_Cubic,YtestCubic)
One_sigma = [1 if \
GPR_pred[i]-GPR_std[i] < YtestGPR[i] < GPR_pred[i]+GPR_std[i] \
else 0 for i in range(len(GPR_pred))]
Two_sigma = [1 if \
GPR_pred[i]-2*GPR_std[i] < YtestGPR[i] < GPR_pred[i]+2*GPR_std[i] \
else 0 for i in range(len(GPR_pred))]
f_1sigma = (sum(One_sigma)/len(One_sigma))*100
f_2sigma = (sum(Two_sigma)/len(Two_sigma))*100
# =========================================================================
# Print Summary
# =========================================================================
print('Mean Y = ',np.mean(Ys))
print('MAE (Cubic | GPR) = ','{:.3e}'.format(Cubic_Errors[1]),\
'|','{:.3e}'.format(GPR_Errors[1]))
print('MSE (Cubic | GPR) = ','{:.3e}'.format(Cubic_Errors[2]),\
'|','{:.3e}'.format(GPR_Errors[2]))
print('RMSE (Cubic | GPR) = ','{:.3e}'.format(Cubic_Errors[3]),\
'|','{:.3e}'.format(GPR_Errors[3]))
print('Mean Rel. Error (%) (Cubic | GPR) = ','{:.3e}'.format(Cubic_Errors[4])\
,'|','{:.3e}'.format(GPR_Errors[4]))
print('Max Rel. Error (%) (Cubic | GPR) = ','{:.3e}'.format(Cubic_Errors[5])\
,'|','{:.3e}'.format(GPR_Errors[5]))
print('R^2 Coeff. of Determination (Cubic | GPR) = ',\
'{:.3e}'.format(RsquareCubic),'|','{:.3e}'.format(RsquareGPR))
Values_Cubic = [error for error in Cubic_Errors]+[RsquareCubic]
Values_GPR = [error for error in GPR_Errors] + [RsquareGPR,
Mean_GPR_std,Max_GPR_std,f_1sigma,f_2sigma]
return Values_Cubic,Values_GPR
def time_smooth_bivariate_spline(self, n_samples):
interpolate.SmoothBivariateSpline(self.x, self.y, self.z)
def extrapolate_emodulus(lut,
datax,
deform,
emod,
deform_norm,
deform_thresh=.05,
inplace=True):
"""Use spline interpolation to fill in nan-values
When points (`datax`, `deform`) are outside the convex
hull of the lut, then :func:`scipy.interpolate.griddata` returns
nan-valules.
With this function, some of these nan-values are extrapolated
using :class:`scipy.interpolate.SmoothBivariateSpline`. The
supported extrapolation values are currently limited to those
where the deformation is above 0.05.
A warning will be issued, because this is not really
recommended.
Parameters
----------
lut: ndarray of shape (N, 3)
The normalized (!! see :func:`normalize`) LUT (first axis is
points, second axis enumerates datax, deform, and emodulus)
datax: ndarray of size N
The normalized x data (corresponding to `lut[:, 0]`)
deform: ndarray of size N
The normalized deform (corresponding to `lut[:, 1]`)
emod: ndarray of size N
The emodulus (corresponding to `lut[:, 2]`); If `emod`
does not contain nan-values, there is nothing to do here.
deform_norm: float
The normalization value used to normalize `lut[:, 1]` and
`deform`.
deform_thresh: float
Not the entire LUT is used for bivariate spline interpolation.
Only the points where `lut[:, 1] > deform_thresh/deform_norm`
are used. This is necessary, because for small deformations,
the LUT has an extreme slope that kills any meaningful
spline interpolation.
inplace: bool
If True (default), replaces nan values in `emod` in-place.
If False, `emod` is not modified.
"""
if not inplace:
emod = np.array(emod, copy=True)
# unknowns are nans and deformation values above the threshold
unkn = np.logical_and(np.isnan(emod), deform > deform_thresh / deform_norm)
if np.sum(unkn) == 0:
# nothing to do
return emod
warnings.warn(
"LUT extrapolation is barely tested and may yield " +
"unphysical values!", KnowWhatYouAreDoingWarning)
lut_crop = lut[lut[:, 1] > deform_thresh / deform_norm, :]
itp = spint.SmoothBivariateSpline(
x=lut_crop[:, 0],
y=lut_crop[:, 1],
z=lut_crop[:, 2],
)
emod[unkn] = itp.ev(datax[unkn], deform[unkn])
return emod
default='mean'
)
args = parser.parse_args()
kind = args.kind
start = time.time()
temp_elev_lst = fs.read_all_tsi_val_temp_elev(VAL, tsi_path)
gridded = fs.populate_temp_elev_dict(temp_elev_lst, grid_temp, grid_elev, single_value=True)
fig = plt.figure()
fig.set_size_inches(14, 14)
grid_x, grid_y = np.mgrid[min(gridded['temp']):max(gridded['temp']):200j,
min(gridded['elev']):max(gridded['elev']):200j]
spline = interpolate.SmoothBivariateSpline(gridded['temp'], gridded['elev'], gridded[kind])
print('knots\n', spline.get_knots())
print('coefficients\n', spline.get_coeffs())
grid_z = spline.ev(grid_x, grid_y)
i = len(temp_elev_lst['val'])
"""
grid_z2 = []
for u in np.linspace(min(gridded['temp']), max(gridded['temp']), 200):
for v in np.linspace(min(gridded['elev']), max(gridded['elev']), 200):
grid_z2.append(eval_surf_spline(u, v, knots[0], knots[1], coeffs))
grid_z2 = np.reshape(grid_z2, (200, 200))
grid_x2, grid_y2 = np.mgrid[min(gridded['temp']):max(gridded['temp']):200j,
def make_Tau(self):
tau_ary = np.load("tau_ary.npy")
self.Tau_ip = ip.SmoothBivariateSpline(tau_ary[0], tau_ary[2],
tau_ary[1])
def sim(S1 = params["sol1nI"],
S1CH = params["sol1cH"],
S1CV = params["sol1cV"],
S2 = params["sol2nI"],
S2CH = params["sol2cH"],
S2CV = params["sol2cV"],
S3 = params["soltnI"],
S3CH = params["soltcH"],
S3CV = params["soltcV"],
H1 = params["hex1G"],
S4 = params["csol1nI"],
S4CH = params["csol1cH"],
S4CV = params["csol1cV"],
S5 = params["csol2nI"],
S5CH = params["csol2cH"],
S5CV = params["csol2cV"],
H2 = params["hex2G"],
S6 = params["csol3nI"],
S6CH = params["csol3cH"],
S6CV = params["csol3cV"],
S7 = params["csol4nI"],
S7CH = params["csol4cH"],
S7CV = params["csol4cV"],
Obj = params["sol3nI"],
ObjCH = params["sol3cH"],
ObjCV = params["sol3cV"],
S9 = params["sol4nI"],
alpha = params["alpha"],
theta = params["theta"],
delta = params["delta"],
seed = 0,
erL = errorsigmaL,
erTh = errorsigmaTheta):
np.random.seed(seed = seed)
rs = [np.random.normal(size = 6) for dummy in range(0, len(eleprefix))]
errors = [[r[0]*erL, r[1]*erL, r[2]*erL,
cos(r[3]*erTh)*cos(r[5]*erTh) - cos(r[4]*erTh)*sin(r[3]*erTh)*sin(r[5]*erTh),
-cos(r[3]*erTh)*sin(r[5]*erTh) - cos(r[4]*erTh)*cos(r[5]*erTh)*sin(r[3]*erTh),
sin(r[3]*erTh)*sin(r[4]*erTh),
cos(r[5]*erTh)*sin(r[3]*erTh) + cos(r[3]*erTh)*cos(r[4]*erTh)*sin(r[5]*erTh),
cos(r[3]*erTh)*cos(r[4]*erTh)*cos(r[5]*erTh) - sin(r[3]*erTh)*sin(r[5]*erTh),
-cos(r[3]*erTh)*sin(r[4]*erTh)] for r in rs]
cmdA = "{} -o {} hexuscope.in {}{}".format(EXE, GDFFILE,
"".join(["{}={} ".format(x,y) for x, y in zip(params.keys(),
[S1, S1CH, S2CV, S2, S2CH, S2CV, H1, S3,
S3CH, S3CV, S4, S4CH, S4CV,
S5, S5CH, S5CV, H2, S6, S6CH, S6CV, S7, S7CH, S7CV, Obj, ObjCH, ObjCV, S9, alpha, theta, delta])]),
"".join(["{}={} ".format(s, t) for x, y in zip(errornames, errors) for s, t in zip(x, y)]))
cmdC = "{} -o {} {} time x y z G".format(EXETRANS, TRANSFILE, GDFFILE)
cmdB = "{} -o {} {}".format(EXETXT, ASCIIFILE, GDFFILE)
cmdD = "{} -o {} {}".format(EXETXT, TRANSASCII, TRANSFILE)
# cmdA,C,D to track the particles, cmdA,B to run standard screen
os.system(cmdA)
# os.system(cmdC)
os.system(cmdB)
# os.system(cmdD)
screen = np.loadtxt(ASCIIFILE, skiprows=5)
x = screen[:,0]
y = screen[:,1]
kx = np.divide(screen[:,4], screen[:,6])
ky = np.divide(screen[:,5], screen[:,6])
meankx = np.mean(kx)
sigkx = np.std(kx)
meanky = np.mean(ky)
sigky = np.std(ky)
N = 40
# set a fixed kx, ky limit if necessary
sigkx = 0.040 / maxsig
sigky = 0.040 / maxsig
x_bins = [[[] for n in range(0,N)] for m in range(0,N)]
y_bins = [[[] for n in range(0,N)] for m in range(0,N)]
x_grid = np.zeros([N, N])
y_grid = np.zeros([N, N])
kx_grid, ky_grid = np.meshgrid(sigkx*np.linspace(-maxsig, maxsig, N),
sigky*np.linspace(-maxsig, maxsig, N))
for xi, yi, kxi, kyi in zip(x, y, kx, ky):
i = int(0.5*N*((kyi-meanky)/(maxsig*sigky)) + 0.5*N)
j = int(0.5*N*((kxi-meankx)/(maxsig*sigkx)) + 0.5*N)
if i < 0 or i > N-1 or j < 0 or j > N-1:
continue
x_bins[i][j].append(xi)
y_bins[i][j].append(yi)
for i in range(0, N):
for j in range(0, N):
x_grid[i,j] = np.mean(x_bins[i][j])
y_grid[i,j] = np.mean(y_bins[i][j])
# Remove possible nan points that would make following interpolation step fail
y_grid[np.isnan(y_grid)]=0
x_grid[np.isnan(x_grid)]=0
index = np.where(x_grid != 0)
xfunc = interpolate.SmoothBivariateSpline(kx_grid[index].flatten(), ky_grid[index].flatten(), x_grid[index].flatten(), kx=5, ky=5)
yfunc = interpolate.SmoothBivariateSpline(kx_grid[index].flatten(), ky_grid[index].flatten(), y_grid[index].flatten(), kx=5, ky=5)
ky_fine = np.linspace(-sigkx*maxsig, sigkx*maxsig, 201)
kx_fine = np.linspace(-sigkx*maxsig, sigkx*maxsig, 201)
FILENAME = "trnsmssn_realpotential.pickle"
FILENAME = 'trnsmssn_new.pickle'
FILENAME = "/home/chenyu/Desktop/GaussianProcess/GPTrelated/trnsmssn_antialiasing.pickle"
with open(FILENAME, "rb") as f:
trnsmssn = pickle.load(f)
shadow = np.array([[trnsmssn((xfunc(kx, ky)[0][0]/sampleScale + sampleL/2)%sampleL,
(yfunc(kx, ky)[0][0]/sampleScale + sampleL/2)%sampleL)[0] for kx in kx_fine] for ky in ky_fine])
return maxsig*sigkx, maxsig*sigky, shadow
smooths = {}
for band in bands:
smooths[band] = 10
for band in bands:
z = ravel(array(z_mat[band]))
ez = ravel(array(ez_mat[band]))
print "fitting band ", band
print inter.__file__
try:
print x.shape
print y.shape
print z.shape
#tck[band] = inter.bisplrep(x, y, z, w=1.0/ez, kx=1, ky=1, s=0.0*len(x))
tck[band] = inter.SmoothBivariateSpline(x,
y,
z,
w=1.0 / ez,
s=smooths[band] * len(x))
except:
print x, y, z, ez, smooths[band] * len(x)
print type(x), x.shape
print type(y), y.shape
print type(z), z.shape
print type(ez), ez.shape
sys.exit(1)
tck["e_" + band] = inter.SmoothBivariateSpline(x,
y,
ez,
kx=1,
ky=1,
s=0.0 * len(x))