def Dn(x, y, Np, ndiv=1, axis=-1, mode='strip', cval=0.):
""" central numerical derivative using Np points of order ndiv
(Np>1 and odd), using convolution
Data needs to be equally spaced in x
can use mode= 'nearest', 'wrap', 'reflect', 'constant'
'strip'
'strip' will just cut the bad data at the ends
cval is for 'constant'
returns x', d^n y/dx^n(x')
Note the algorithm is not intended to remove noise
But to provide more accurate derivative of a function.
The larger Np the more deriviatives are available.
It basically finds the best taylor series parameter
assuming Np around the center are available:
assuming f_k = f(xo + k dx), k=-n .. n, Np=2*n+1
and with f(x) = f(xo) + f' (x-xo) + f''(x-xo)^2/2 + ....
= f(xo) + f' k dx + ((f'' dx**2)/2) k**2 + ...
we want to solve for (1, f', f'' dx**2/2, ...)
and we pick the answer for the correct derrivative.
"""
dx = x[1] - x[0]
kernel = central_diff_weights(Np, ndiv=ndiv)
strip = False
if mode == 'strip':
strip = True
mode = 'reflect'
dy = filters.correlate1d(y, kernel, axis=axis, mode=mode, cval=cval)
D = dy / dx**ndiv
if strip:
x, D = _do_strip(x, D, Np, axis=axis)
return x, D
def coef_restriction_diffseq(n_coeffs,
degree=1,
n_vars=None,
position=0,
base_idx=0):
#check boundaries, returns "valid" ?
if degree == 1:
diff_coeffs = [-1, 1]
n_points = 2
elif degree > 1:
from scipy import misc
n_points = next_odd(degree + 1) #next odd integer after degree+1
diff_coeffs = misc.central_diff_weights(n_points, ndiv=degree)
dff = np.concatenate((diff_coeffs, np.zeros(n_coeffs - len(diff_coeffs))))
from scipy import linalg
reduced = linalg.toeplitz(dff, np.zeros(n_coeffs - len(diff_coeffs) + 1)).T
#reduced = np.kron(np.eye(n_coeffs-n_points), diff_coeffs)
if n_vars is None:
return reduced
else:
full = np.zeros((n_coeffs - 1, n_vars))
full[:, position:position + n_coeffs] = reduced
return full
def Dn(x, y, Np, ndiv=1, axis=-1, mode='strip', cval=0.):
""" central numerical derivative using Np points of order ndiv
(Np>1 and odd), using convolution
Data needs to be equally spaced in x
can use mode= 'nearest', 'wrap', 'reflect', 'constant'
'strip'
'strip' will just cut the bad data at the ends
cval is for 'constant'
returns x', d^n y/dx^n(x')
Note the algorithm is not intended to remove noise
But to provide more accurate derivative of a function.
The larger Np the more deriviatives are available.
It basically finds the best taylor series parameter
assuming Np around the center are available:
assuming f_k = f(xo + k dx), k=-n .. n, Np=2*n+1
and with f(x) = f(xo) + f' (x-xo) + f''(x-xo)^2/2 + ....
= f(xo) + f' k dx + ((f'' dx**2)/2) k**2 + ...
we want to solve for (1, f', f'' dx**2/2, ...)
and we pick the answer for the correct derrivative.
"""
dx = x[1]-x[0]
kernel = central_diff_weights(Np,ndiv=ndiv)
strip = False
if mode=='strip':
strip = True
mode = 'reflect'
dy = filters.correlate1d(y, kernel, axis=axis, mode=mode, cval=cval)
D = dy/dx**ndiv
if strip:
x, D = _do_strip(x, D, Np, axis=axis)
return x, D
def test_get_diff_coeff(self):
forward_11 = get_diff_coeff([0, 1], 1)
forward_13 = get_diff_coeff([0, 1, 2, 3], 1)
backward_26 = get_diff_coeff(np.arange(-6, 1), 2)
central_29 = get_diff_coeff(np.arange(-4, 5), 2)
self.assertArrayAlmostEqual(forward_11, [-1, 1])
self.assertArrayAlmostEqual(forward_13, [-11./6, 3, -3./2, 1./3])
self.assertArrayAlmostEqual(backward_26, [137./180, -27./5,33./2,-254./9,
117./4,-87./5,203./45])
self.assertArrayAlmostEqual(central_29, central_diff_weights(9, 2))
def test_get_diff_coeff(self):
forward_11 = get_diff_coeff([0, 1], 1)
forward_13 = get_diff_coeff([0, 1, 2, 3], 1)
backward_26 = get_diff_coeff(np.arange(-6, 1), 2)
central_29 = get_diff_coeff(np.arange(-4, 5), 2)
self.assertArrayAlmostEqual(forward_11, [-1, 1])
self.assertArrayAlmostEqual(forward_13, [-11./6, 3, -3./2, 1./3])
self.assertArrayAlmostEqual(backward_26, [137./180, -27./5,33./2,-254./9,
117./4,-87./5,203./45])
self.assertArrayAlmostEqual(central_29, central_diff_weights(9, 2))
def raytraceX(obj, ps, sys_index, nimgs=None, eps=None):
#---------------------------------------------------------------------------
# Find the derivative of the arrival time surface.
#---------------------------------------------------------------------------
arrival = obj.basis.arrival_grid(ps)[sys_index]
w = central_diff_weights(3)
d = abs(correlate1d(arrival, w, axis=0, mode='constant')) \
+ abs(correlate1d(arrival, w, axis=1, mode='constant'))
d = d[1:-1,1:-1]
pl.matshow(d)
xy = obj.basis.refined_xy_grid(ps)[1:-1,1:-1]
# Create flattened *views*
xy = xy.ravel()
dravel = d.ravel()
imgs = []
offs = []
print 'searching...'
for i in argsort(dravel):
if nimgs == len(imgs): break
new_image = True
for img in imgs:
if abs(img-xy[i]) <= eps:
new_image = False
if new_image:
imgs.append(xy[i])
offs.append(i)
#---------------------------------------------------------------------------
# Print the output
#---------------------------------------------------------------------------
if imgs:
#print imgs
#if len(imgs) % 2 == 1: imgs = imgs[:-1] # XXX: Correct?
imgs = array(imgs)
g0 = array(arrival[1:-1,1:-1], copy=True)
g0ravel = g0.ravel()
times = g0ravel[offs]
order = argsort(times)
else:
order = []
return [(times[i], imgs[i]) for i in order]
def diff_finite(
dem: np.ndarray,
cellsize: tuple,
dx: int,
dy: int,
) -> np.ndarray:
"""
Calculate a higher-order derivative of the DEM by finite differencing
successively in X and then Y direction. Uses a central differencing scheme.
:param dem: DEM (z values at each pixel of a rectangular grid)
:param cellsize: pixel ``(width, height)``
:param dx: order of derivative in x direction
:param dy: order of derivative in y direction
"""
wx = misc.central_diff_weights(dx + 1 + dx % 2, dx).reshape(1, -1)
wy = misc.central_diff_weights(dy + 1 + dy % 2, dy).reshape(-1, 1)
image = _as_float_array(dem)
image = ndimage.convolve(image, wx, mode='nearest')
image = ndimage.convolve(image, wy, mode='nearest')
norm = np.product(cellsize**np.array([dy, dx]))
return image / norm
def _get_central_diff_weights(order):
"""Determine the weights for central differentiation"""
if order == 3:
weights = np.array([-1, 0, 1]) / 2.0
elif order == 5:
weights = np.array([1, -8, 0, 8, -1]) / 12.0
elif order == 7:
weights = np.array([-1, 9, -45, 0, 45, -9, 1]) / 60.0
elif order == 9:
weights = np.array([3, -32, 168, -672, 0, 672, -168, 32, -3]) / 840.0
else:
weights = central_diff_weights(order, 1)
return weights
def n_order_scheme(f, x, n, h):
"""
n-order differentiable scheme
"""
weights = central_diff_weights(n)
grad_f = 0
i_s, j_s = list(zip(*list(enumerate(np.arange(-n // 2, n // 2) + 1))))
if PARALLEL:
f_js = Parallel(n_jobs=len(j_s), prefer="threads")(delayed(f)(y=x + j * h) for j in j_s)
else:
f_js = [f(y=x + j * h) for j in j_s]
for i, f_j in zip(i_s, f_js):
if weights[i] != 0:
grad_f += weights[i] * f_j
return grad_f / h
def deriv(env, model, obj_index, src_index, m, axis, R):
w = central_diff_weights(5)
#d = correlate1d(m, w, axis=axis, mode='constant')
d = (correlate1d(m, -w, axis=0, mode='constant')) \
+ (correlate1d(m, w, axis=1, mode='constant'))
d = (correlate1d(d, -w, axis=0, mode='constant')) \
+ (correlate1d(d, w, axis=1, mode='constant'))
d = d[2:-2, 2:-2]
d[d > .8] = .8
d[d < -.8] = -.8
#d = correlate1d(d, w, axis=axis, mode='constant')
#d = diff(d, axis=axis)
#d /= abs(d)
#d = correlate1d(d, w, axis=axis, mode='constant')
#d = diff(d, axis=axis)
R -= model[0].basis.top_level_cell_size * 2
#R -= model[0].basis.top_level_cell_size * 2
pl.matshow(d, extent=[-R, R, -R, R])
glspl.colorbar()
arrival_plot(model, obj_index, src_index, only_contours=True, clevels=200)
def deriv(env, model, obj_index, src_index, m, axis, R):
w = central_diff_weights(5)
#d = correlate1d(m, w, axis=axis, mode='constant')
d = (correlate1d(m, -w, axis=0, mode='constant')) \
+ (correlate1d(m, w, axis=1, mode='constant'))
d = (correlate1d(d, -w, axis=0, mode='constant')) \
+ (correlate1d(d, w, axis=1, mode='constant'))
d = d[2:-2,2:-2]
d[d>.8] = .8
d[d<-.8] = -.8
#d = correlate1d(d, w, axis=axis, mode='constant')
#d = diff(d, axis=axis)
#d /= abs(d)
#d = correlate1d(d, w, axis=axis, mode='constant')
#d = diff(d, axis=axis)
R -= model[0].basis.top_level_cell_size * 2
#R -= model[0].basis.top_level_cell_size * 2
pl.matshow(d, extent=[-R,R,-R,R])
glspl.colorbar()
arrival_plot(model, obj_index, src_index, only_contours=True, clevels=200)
def coef_restriction_diffseq(n_coeffs, degree=1, n_vars=None, position=0, base_idx=0):
#check boundaries, returns "valid" ?
if degree == 1:
diff_coeffs = [-1, 1]
n_points = 2
elif degree > 1:
from scipy import misc
n_points = next_odd(degree + 1) #next odd integer after degree+1
diff_coeffs = misc.central_diff_weights(n_points, ndiv=degree)
dff = np.concatenate((diff_coeffs, np.zeros(n_coeffs - len(diff_coeffs))))
from scipy import linalg
reduced = linalg.toeplitz(dff, np.zeros(n_coeffs - len(diff_coeffs) + 1)).T
#reduced = np.kron(np.eye(n_coeffs-n_points), diff_coeffs)
if n_vars is None:
return reduced
else:
full = np.zeros((n_coeffs-1, n_vars))
full[:, position:position+n_coeffs] = reduced
return full
def grad_tau(env, model, obj_index, which, src_index):
assert which in ['x','y'], "grad_tau: 'which' must be one of 'x' or 'y'"
#print "grad_tau"
obj,ps = model['obj,data'][obj_index]
R = obj.basis.mapextent
#---------------------------------------------------------------------------
# Find the derivative of the arrival time surface.
#---------------------------------------------------------------------------
arrival = obj.basis.arrival_grid(ps)[src_index]
w = central_diff_weights(3)
which = 1 if which == 'x' else 0
d = correlate1d(arrival, w, axis=which, mode='constant')
d = d[1:-1,1:-1]
d[np.abs(d) < 1e-3] = 0
d[d>0] = 1
d[d<0] = -1
pl.matshow(d, fignum=False, extent=[-R,R,-R,R], alpha=0.5)
def grad_tau(env, model, obj_index, which, src_index):
assert which in ['x', 'y'], "grad_tau: 'which' must be one of 'x' or 'y'"
#print "grad_tau"
obj, ps = model['obj,data'][obj_index]
R = obj.basis.mapextent
#---------------------------------------------------------------------------
# Find the derivative of the arrival time surface.
#---------------------------------------------------------------------------
arrival = obj.basis.arrival_grid(ps)[src_index]
w = central_diff_weights(3)
which = 1 if which == 'x' else 0
d = correlate1d(arrival, w, axis=which, mode='constant')
d = d[1:-1, 1:-1]
d[np.abs(d) < 1e-3] = 0
d[d > 0] = 1
d[d < 0] = -1
pl.matshow(d, fignum=False, extent=[-R, R, -R, R], alpha=0.5)
def gradient(derivative, points, shape, step=None, order='C'):
''' Approximate the derivative with a central difference.
Parameters
----------
derivative : int
The order of the derivative to approximate.
points : int
The number of points in the central difference. Must be an odd
integer greater than `derivative`.
shape : tuple
The shape of the array in non-vector form.
step: tuple, optional
The step sizes between adjacent values along each dimension in the
array. Passing None defaults to step size 1.0 along all dimensions.
order = {'C', 'F', 'A'}, optional
The order by which the vectorized array is reshaped. This is the
same parameter as given to functions like numpy.reshape. For a
discussion of the memory efficiency of different orders and how that
is determined by the underlying format, see the documentation of
commands that take an order argument.
Returns
-------
LinearOperator
A LinearOperator implementing the central difference on vectorized
inputs.
Raises
------
ValueError
When the inputs are not the right dimensions.
See Also
--------
convolve : The central difference is calculated using this LinearOperator.
scipy.misc.central_diff_weights : The scipy function returning the
required weights for calculating the central difference.
Examples
--------
>>> import numpy as np
>>> from pyop.operators import gradient
>>> grad = gradient(1, 3, (5,6))
>>> laplace = gradient(2, 3, (5,6), step=(0.5,2))
>>> A = np.indices((5,6)).sum(0)**2 / 2.
>>> grad(np.ravel(A)).reshape(5,6)
array([[ 0.5 , 2. , 4.25, 7. , 10.25, 5. ],
[ 2. , 4. , 6. , 8. , 10. , -0.25],
[ 4.25, 6. , 8. , 10. , 12. , -2. ],
[ 7. , 8. , 10. , 12. , 14. , -4.25],
[ 4. , 1. , -0.25, -2. , -4.25, -32. ]])
>>> laplace(np.ravel(A)).reshape(5,6)
array([[ 2.125, 4.25 , 2.25 , -3.75 , -13.75 , -32.25 ],
[ 4.25 , 4.25 , 4.25 , 4.25 , 4.25 , -1.875],
[ 4.125, 4.25 , 4.25 , 4.25 , 4.25 , -3.75 ],
[ 3.75 , 4.25 , 4.25 , 4.25 , 4.25 , -5.875],
[ -46.875, -67.75 , -93.75 , -123.75 , -157.75 , -208.25 ]])
'''
# it stops being useful at all at this point
if any(s < points for s in shape):
raise ValueError("Shape's dims must have at least as many "
"points as central difference weights.")
if step is None:
step = (1.0, ) * len(shape)
elif len(step) != len(shape):
raise ValueError("Shape and step must have same ndims (length).")
step = enumerate(float(s) ** derivative for s in step)
# reverses the order of the weights to compensate for convolution's
# "flipped" shift
weights = central_diff_weights(points, derivative)[::-1]
# create a kernel with ndim specified by shape, each of length points
kernel = np.zeros((points, ) * len(shape))
# slice that gets the center values along first dimension
slc = (slice(None), ) + (points // 2, ) * (kernel.ndim - 1)
# fill in kernel along the center of each dimension
for dim, s in step:
kernel[slc[-dim:] + slc[:-dim]] += weights / s
return convolve(kernel, shape, order)
import copy
import numpy as np
import math
from scipy.misc import central_diff_weights
import matplotlib.pyplot as plt
from scipy.fftpack import fft
from scipy.io import wavfile # get the api
from scipy.interpolate import UnivariateSpline
fs, data = wavfile.read("c:/temp/06DieWasserflut.wav") # load the data
a = data.T[0] # this is a two channel soundtrack, I get the first track
b = [(ele / 2**16.) * 2 - 1
for ele in a] # this is 16-bit track, b is now normalized on [-1,1)
i = 0
w1 = central_diff_weights(49, 1)
step = 10240
def freq2scale(freq):
return 69 + 12 * math.log(freq, 2)
print(fs)
print("===========")
print("time in ms:intense:freq:name")
plot_x = []
plot_y = []
plot_count = 0
while i < len(a) - step:
# deriv.py: 数値微分
import scipy.misc as scmisc
import numpy as np
from tktools import relerr # 相対誤差
# 中央差分式の係数(奇数点数のみ)を出力
for m in range(1, 6):
print('--- ', m, '階差分 ---')
if m >= 3:
min_num_points = 2 * m - 1
else:
min_num_points = 3
for n in range(min_num_points, 10, 2):
print(n, '点公式: ', scmisc.central_diff_weights(n, m))
# 元の関数
def org_func(x):
return np.sin(np.cos(x))
# 1階導関数
def diff_func(x):
return np.cos(np.cos(x)) * (-np.sin(x))
# 中心差分商に基づく数値微分
x = 1.5
print('真値 = ', diff_func(x))
b = funcion.diff(x,1,y,2)
c = S.Derivative(funcion,z,3)
c.doit()
import numpy as N
import scipy.misc as SM
#Número de puntos ha de ser impar
x = S.symbols('x',real=True)
funsym = S.exp(-x**2)
funnum = S.lambdify(x,funsym,"numpy")
z= 2
solsym = funsym.diff(x,1).subs(x,z).evalf()
solnum = SM.derivative(funnum,z, dx=1e-6)
pesos = SM.central_diff_weights(3)
print(solnum,pesos)
xs = S.symbols('x',real=True)
fsym = S.exp(-xs**2)
fnum = S.lambdify(xs,fsym,"numpy")
x = N.arange(-1,1,1/4)
y = fnum(x)
derfsym = fsym.diff(xs,1)
solsym = list([derfsym.subs(xs,z)for z in x])
s0 = N.array(solsym).astype(float)
s1 = N.diff(y)/N.diff(x)
s2 = N.gradient(y,x)
s3 = N.gradient(y,x,edge_order=2)
def toec_fit(strains, stresses, eq_stress = None, zero_crit=1e-10):
"""
A third-order elastic constant fitting function based on
central-difference derivatives with respect to distinct
strain states. The algorithm is summarized as follows:
1. Identify distinct strain states as sets of indices
for which nonzero strain values exist, typically
[(0), (1), (2), (3), (4), (5), (0, 1) etc.]
2. For each strain state, find and sort strains and
stresses by strain value.
3. Find first and second derivatives of each stress
with respect to scalar variable corresponding to
the smallest perturbation in the strain.
4. Use the pseudoinverse of a matrix-vector expression
corresponding to the parameterized stress-strain
relationship and multiply that matrix by the respective
calculated first or second derivatives from the
previous step.
5. Place the calculated second and third-order elastic
constants appropriately.
Args:
strains (nx3x3 array-like): Array of 3x3 strains
to use in fitting of TOEC and SOEC
stresses (nx3x3 array-like): Array of 3x3 stresses
to use in fitting of TOEC and SOEC. These
should be PK2 stresses.
eq_stress (3x3 array-like): stress corresponding to
equilibrium strain (i. e. "0" strain state).
If not specified, function will try to find
the state in the list of provided stresses
and strains. If not found, defaults to 0.
zero_crit (float): value for which strains below
are ignored in identifying strain states.
"""
if len(stresses) != len(strains):
raise ValueError("Length of strains and stresses are not equivalent")
vstresses = np.array([Stress(stress).voigt for stress in stresses])
vstrains = np.array([Strain(strain).voigt for strain in strains])
vstrains[np.abs(vstrains) < zero_crit] = 0
# Try to find eq_stress if not specified
if eq_stress is not None:
veq_stress = Stress(eq_stress).voigt
else:
veq_stress = vstresses[np.all(vstrains==0, axis=1)]
if veq_stress:
if np.shape(veq_stress) > 1 and not \
(abs(veq_stress - veq_stress[0]) < 1e-8).all():
raise ValueError("Multiple stresses found for equilibrium strain"
" state, please specify equilibrium stress or "
" remove extraneous stresses.")
veq_stress = veq_stress[0]
else:
veq_stress = np.zeros(6)
# Collect independent strain states:
independent = set([tuple(np.nonzero(vstrain)[0].tolist())
for vstrain in vstrains])
strain_states = []
dsde = np.zeros((6, len(independent)))
d2sde2 = np.zeros((6, len(independent)))
for n, ind in enumerate(independent):
# match strains with templates
template = np.zeros(6, dtype=bool)
np.put(template, ind, True)
template = np.tile(template, [vstresses.shape[0], 1])
mode = (template == (np.abs(vstrains) > 1e-10)).all(axis=1)
mstresses = vstresses[mode]
mstrains = vstrains[mode]
# add zero strain state
mstrains = np.vstack([mstrains, np.zeros(6)])
mstresses = np.vstack([mstresses, np.zeros(6)])
# sort strains/stresses by strain values
mstresses = mstresses[mstrains[:, ind[0]].argsort()]
mstrains = mstrains[mstrains[:, ind[0]].argsort()]
strain_states.append(mstrains[-1] / \
np.min(mstrains[-1][np.nonzero(mstrains[0])]))
diff = np.diff(mstrains, axis=0)
if not (abs(diff - diff[0]) < 1e-8).all():
raise ValueError("Stencil for strain state {} must be odd-sampling"
" centered at 0.".format(ind))
h = np.min(diff[np.nonzero(diff)])
coef1 = central_diff_weights(len(mstresses), 1)
coef2 = central_diff_weights(len(mstresses), 2)
if eq_stress is not None:
mstresses[len(mstresses) // 2] = veq_stress
dsde[:, n] = np.dot(np.transpose(mstresses), coef1) / h
d2sde2[:, n] = np.dot(np.transpose(mstresses), coef2) / h**2
m2i, m3i = generate_pseudo(strain_states)
s2vec = np.ravel(dsde.T)
c2vec = np.dot(m2i, s2vec)
c2 = np.zeros((6, 6))
c2[np.triu_indices(6)] = c2vec
c2 = c2 + c2.T - np.diag(np.diag(c2))
c3 = np.zeros((6, 6, 6))
s3vec = np.ravel(d2sde2.T)
c3vec = np.dot(m3i, s3vec)
list_indices = list(itertools.combinations_with_replacement(range(6), r=3))
indices_ij = itertools.combinations_with_replacement(range(6), r=3)
indices = list(itertools.combinations_with_replacement(range(6), r=3))
for n, (i, j, k) in enumerate(indices):
c3[i,j,k] = c3[i,k,j] = c3[j,i,k] = c3[j,k,i] = \
c3[k,i,j] = c3[k,j,i] = c3vec[n]
return TensorBase.from_voigt(c2), TensorBase.from_voigt(c3)