Goal
I would like to compute the 3D volume integral of a numeric scalar field.
Code
For this post, I will use an example of which the integral can be exactly computed. I have therefore chosen the following function:
In Python, I define the function, and a set of points in 3D, and then generate the discrete values at these points:
import numpy as np
# Make data.
def function(x, y, z):
return x**y**z
N = 5
grid = np.meshgrid(
np.linspace(0, 1, N),
np.linspace(0, 1, N),
np.linspace(0, 1, N)
)
points = np.vstack(list(map(np.ravel, grid))).T
x = points[:, 0]
y = points[:, 1]
z = points[:, 2]
values = [function(points[i, 0], points[i, 1], points[i, 2])
for i in range(len(points))]
Question
How can I find the integral, if I don't know the underlying function, i.e. if I only have the coordinates (x, y, z) and the values?
A nice way to go about this would be using scipy's tplquad integration. However, to use that, we need a function and not a cloud point.
An easy way around that is to use an interpolator, to get a function approximating our cloud point - we can for example use scipy's RegularGridInterpolator if the data is on a regular grid:
import numpy as np
from scipy import integrate
from scipy.interpolate import RegularGridInterpolator
# Make data.
def function(x,y,z):
return x*y*z
N = 5
xmin, xmax = 0, 1
ymin, ymax = 0, 1
zmin, zmax = 0, 1
x = np.linspace(xmin, xmax, N)
y = np.linspace(ymin, ymax, N)
z = np.linspace(zmin, zmax, N)
values = function(*np.meshgrid(x,y,z, indexing='ij'))
# Interpolate:
function_interpolated = RegularGridInterpolator((x, y, z), values)
# tplquad integrates func(z,y,x)
f = lambda z,y,x : my_interpolating_function([z,y,x])
result, error = integrate.tplquad(f, xmin, xmax, lambda _: ymin, lambda _:ymax,lambda *_: zmin, lambda *_: zmax)
In the example above, we get result = 0.12499999999999999 - close enough!
The easiest way to achieve what you are looking for is probably scipy's integration function. Here your example:
from scipy import integrate
# Make data.
def func(x,y,z):
return x**y**z
ranges = [[0,1], [0,1], [0,1]]
result, error = integrate.nquad(func, ranges)
Are you aware that the function that you created is different from the one that you show in the image. The one you created is an exponential (x^y^z) while the one that you are showing is just multiplications. If you want to represent the function in the image, use
def func(x,y,z):
return x*y*z
Hope this answers your question, otherwise just write a comment!
Edit:
Misread your post. If you only have the results, and they are not regularly spaced, you would have to figure out some form of interpolation (i.e. linear) and a lookup-table. If you do not know how to create that, let me know. The rest of the stated answer could still be used if you define func to return interpolated values from your original data
The first answer explains nicely the principal approach to handle this. Just wanted to illustrate an alternative way by showing the power of sklearn package and machine learning regression.
Doing the meshgrid in 3D gives a very large numpy array,
import numpy as np
N = 5
xmin, xmax = 0, 1
ymin, ymax = 0, 1
zmin, zmax = 0, 1
x = np.linspace(xmin, xmax, N)
y = np.linspace(ymin, ymax, N)
z = np.linspace(zmin, zmax, N)
grid = np.array(np.meshgrid(x,y,z, indexing='ij'))
grid.shape = (3, 5, 5, 5) # 2*5*5*5 = 250 numbers
Which is visually not very intuitive with 250 numbers. With different possible indexing ('ij' or 'xy'). Using regression we can get the same result with few input points (15-20).
# building random combinations from (x,y,z)
X = np.random.choice(x, 20)[:,None]
Y = np.random.choice(y, 20)[:,None]
Z = np.random.choice(z, 20)[:,None]
xyz = np.concatenate((X,Y,Z), axis = 1)
data = np.multiply.reduce(xyz, axis = 1)
So the input (grid) is just a 2D numpy array,
xyz.shape
(20, 3)
With the corresponding data,
data.shape = (20,)
Now the regression function and integration,
from sklearn.preprocessing import PolynomialFeatures
from sklearn.linear_model import LinearRegression
from sklearn.pipeline import Pipeline
from scipy import integrate
pipe=Pipeline([('polynomial',PolynomialFeatures(degree=3)),('modal',LinearRegression())])
pipe.fit(xyz, data)
def func(x,y,z):
return pipe.predict([[x, y, z]])
ranges = [[0,1], [0,1], [0,1]]
result, error = integrate.nquad(func, ranges)
print(result)
0.1257
This approach is useful with limited number of points.
Based on your requirements, it sounds like the most appropriate technique would be Monte Carlo integration:
# Step 0 start with some empirical data
observed_points = np.random.uniform(0,1,size=(10000,3))
unknown_fn = lambda x: np.prod(x) # just used to generate fake values
observed_values = np.apply_along_axis(unknown_fn, 1, observed_points)
K = 1000000
# Step 1 - assume that f(x,y,z) can be approximated by an interpolation
# of the data we have (you could get really fancy with the
# selection of interpolation method - we'll stick with straight lines here)
from scipy.interpolate import LinearNDInterpolator
f_interpolate = LinearNDInterpolator(observed_points, observed_values)
# Step 2 randomly sample from within convex hull of observed data
# Step 2a - Uniformly sample from bounding 3D-box of data
lower_bounds = observed_points.min(axis=0)
upper_bounds = observed_points.max(axis=0)
sampled_points = np.random.uniform(lower_bounds, upper_bounds,size=(K, 3))
# Step 2b - Reject points outside of convex hull...
# Luckily, we get a np.nan from LinearNDInterpolator in this case
sampled_values = f_interpolate(sampled_points)
rejected_idxs = np.argwhere(np.isnan(sampled_values))
# Step 2c - Remember accepted values of estimated f(x_i, y_i, z_i)
final_sampled_values = np.delete(sampled_values, rejected_idxs, axis=0)
# Step 3 - Calculate estimate of volume of observed data domain
# Since we sampled uniformly from the convex hull of data domain,
# each point was selected with P(x,y,z)= 1 / Volume of convex hull
volume = scipy.spatial.ConvexHull(observed_points).volume
# Step 4 - Multiply estimated volume of domain by average sampled value
I_hat = volume * final_sampled_values.mean()
print(I_hat)
For a derivation of why this works see this: https://cs.dartmouth.edu/wjarosz/publications/dissertation/appendixA.pdf
Related
I'm trying to generate a regular n number of points within the volume of a sphere. I found this similar answer (https://scicomp.stackexchange.com/questions/29959/uniform-dots-distribution-in-a-sphere) on generating a uniform regular n number of points on the surface of a sphere, with the following code:
import numpy as np
n = 5000
r = 1
z = []
y = []
x = []
alpha = 4.0*np.pi*r*r/n
d = np.sqrt(alpha)
m_nu = int(np.round(np.pi/d))
d_nu = np.pi/m_nu
d_phi = alpha/d_nu
count = 0
for m in range (0,m_nu):
nu = np.pi*(m+0.5)/m_nu
m_phi = int(np.round(2*np.pi*np.sin(nu)/d_phi))
for n in range (0,m_phi):
phi = 2*np.pi*n/m_phi
xp = r*np.sin(nu)*np.cos(phi)
yp = r*np.sin(nu)*np.sin(phi)
zp = r*np.cos(nu)
x.append(xp)
y.append(yp)
z.append(zp)
count = count +1
which works as intended:
How can I modify this to generate a regular set of n points in the volume of a sphere?
Another method to do this, yielding uniformity in volume:
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
import numpy as np
dim_len = 30
spacing = 2 / dim_len
point_cloud = np.mgrid[-1:1:spacing, -1:1:spacing, -1:1:spacing].reshape(3, -1).T
point_radius = np.linalg.norm(point_cloud, axis=1)
sphere_radius = 0.5
in_points = point_radius < sphere_radius
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
ax.scatter(point_cloud[in_points, 0], point_cloud[in_points, 1], point_cloud[in_points, 2], )
plt.show()
Output (matplotlib mixes up the view but it is a uniformly sampled sphere (in volume))
Uniform sampling, then checking if points are in the sphere or not by their radius.
Uniform sampling reference [see this answer's edit history for naiive sampling].
This method has the drawback of generating redundant points which are then discarded.
It has the upside of vectorization, which probably makes up for the drawback. I didn't check.
With fancy indexing, one could generate the same points as this method without generating redundant points, but I doubt it can be easily (or at all) vectorized.
Sample uniformly along X. For every value of X, you draw two Y from X²+Y²=1. Sample uniformly between these two Y. Then for every (X, Y) pair, you draw two Z from X²+Y²+Z²=1. Sample uniformly between these two Z.
I am trying to fit a line to the 9.0 to 10.0 um regime of my data set. Here is my plot:
Unfortunately, it's a scatter plot with the x values not being indexed from small numbers to large numbers so I can't just apply the optimize.curve_fit function to a specific range of indices to get the desired range in x values.
Below is my go-to procedure for curve fitting. How would I modify it to only get a fit for the 9.0 to 10.0 um x-value range (in my case, the x_dist variable) which has points scattered randomly throughout the indices?
def func(x,a,b): # Define your fitting function
return a*x+b
initialguess = [-14.0, 0.05] # initial guess for the parameters of the function func
fit, covariance = optimize.curve_fit( # call to the fitting routine curve_fit. Returns optimal values of the fit parameters, and their estimated variance
func, # function to fit
x_dist, # data for independant variable
xdiff_norm, # data for dependant variable
initialguess, # initial guess of fit parameters
) # uncertainty in dependant variable
print("linear coefficient:",fit[0],"+-",np.sqrt(covariance[0][0])) #print value and one std deviation of first fit parameter
print("offset coefficient:",fit[1],"+-",np.sqrt(covariance[1][1])) #print value and one std deviation of second fit parameter
print(covariance)
You correctly identified that the problem arises because your x-value data are not ordered. You can address this problem differently. One way is to use Boolean masks to filter out the unwanted values. I tried to be as close as possible to your example:
from matplotlib import pyplot as plt
import numpy as np
from scipy import optimize
#fake data generation
np.random.seed(1234)
arr = np.linspace(0, 15, 100).reshape(2, 50)
arr[1, :] = np.random.random(50)
arr[1, 20:45] += 2 * arr[0, 20:45] -5
rng = np.random.default_rng()
rng.shuffle(arr, axis = 1)
x_dist = arr[0, :]
xdiff_norm = arr[1, :]
def func(x, a, b):
return a * x + b
initialguess = [5, 3]
mask = (x_dist>2.5) & (x_dist<6.6)
fit, covariance = optimize.curve_fit(
func,
x_dist[mask],
xdiff_norm[mask],
initialguess)
plt.scatter(x_dist, xdiff_norm, label="data")
x_fit = np.linspace(x_dist[mask].min(), x_dist[mask].max(), 100)
y_fit = func(x_fit, *fit)
plt.plot(x_fit, y_fit, c="red", label="fit")
plt.legend()
plt.show()
Sample output:
This approach does not modify x_dist and xdiff_norm which might or might not be a good thing for further data evaluation. If you wanted to use a line plot instead of a scatter plot, it might be rather useful to sort your arrays in advance (try a line plot with the above method to see why):
from matplotlib import pyplot as plt
import numpy as np
from scipy import optimize
#fake data generation
np.random.seed(1234)
arr = np.linspace(0, 15, 100).reshape(2, 50)
arr[1, :] = np.random.random(50)
arr[1, 20:45] += 2 * arr[0, 20:45] -5
rng = np.random.default_rng()
rng.shuffle(arr, axis = 1)
x_dist = arr[0, :]
xdiff_norm = arr[1, :]
def func(x, a, b):
return a * x + b
#find indexes of a sorted x_dist array, then sort both arrays based on this index
ind = x_dist.argsort()
x_dist = x_dist[ind]
xdiff_norm = xdiff_norm[ind]
#identify index where linear range starts for normal array indexing
start = np.argmax(x_dist>2.5)
stop = np.argmax(x_dist>6.6)
initialguess = [5, 3]
fit, covariance = optimize.curve_fit(
func,
x_dist[start:stop],
xdiff_norm[start:stop],
initialguess)
plt.plot(x_dist, xdiff_norm, label="data")
x_fit = np.linspace(x_dist[start], x_dist[stop], 100)
y_fit = func(x_fit, *fit)
plt.plot(x_fit, y_fit, c="red", ls="--", label="fit")
plt.legend()
plt.show()
Sample output (unsurprisingly not much different):
I need to create a function which would be the inverse of the np.gradient function.
Where the Vx,Vy arrays (Velocity component vectors) are the input and the output would be an array of anti-derivatives (Arrival Time) at the datapoints x,y.
I have data on a (x,y) grid with scalar values (time) at each point.
I have used the numpy gradient function and linear interpolation to determine the gradient vector Velocity (Vx,Vy) at each point (See below).
I have achieved this by:
#LinearTriInterpolator applied to a delaunay triangular mesh
LTI= LinearTriInterpolator(masked_triang, time_array)
#Gradient requested at the mesh nodes:
(Vx, Vy) = LTI.gradient(triang.x, triang.y)
The first image below shows the velocity vectors at each point, and the point labels represent the time value which formed the derivatives (Vx,Vy)
The next image shows the resultant scalar value of the derivatives (Vx,Vy) plotted as a colored contour graph with associated node labels.
So my challenge is:
I need to reverse the process!
Using the gradient vectors (Vx,Vy) or the resultant scalar value to determine the original Time-Value at that point.
Is this possible?
Knowing that the numpy.gradient function is computed using second order accurate central differences in the interior points and either first or second order accurate one-sides (forward or backwards) differences at the boundaries, I am sure there is a function which would reverse this process.
I was thinking that taking a line derivative between the original point (t=0 at x1,y1) to any point (xi,yi) over the Vx,Vy plane would give me the sum of the velocity components. I could then divide this value by the distance between the two points to get the time taken..
Would this approach work? And if so, which numpy integrate function would be best applied?
An example of my data can be found here [http://www.filedropper.com/calculatearrivaltimefromgradientvalues060820]
Your help would be greatly appreciated
EDIT:
Maybe this simplified drawing might help understand where I'm trying to get to..
EDIT:
Thanks to #Aguy who has contibuted to this code.. I Have tried to get a more accurate representation using a meshgrid of spacing 0.5 x 0.5m and calculating the gradient at each meshpoint, however I am not able to integrate it properly. I also have some edge affects which are affecting the results that I don't know how to correct.
import numpy as np
from scipy import interpolate
from matplotlib import pyplot
from mpl_toolkits.mplot3d import Axes3D
#Createmesh grid with a spacing of 0.5 x 0.5
stepx = 0.5
stepy = 0.5
xx = np.arange(min(x), max(x), stepx)
yy = np.arange(min(y), max(y), stepy)
xgrid, ygrid = np.meshgrid(xx, yy)
grid_z1 = interpolate.griddata((x,y), Arrival_Time, (xgrid, ygrid), method='linear') #Interpolating the Time values
#Formatdata
X = np.ravel(xgrid)
Y= np.ravel(ygrid)
zs = np.ravel(grid_z1)
Z = zs.reshape(X.shape)
#Calculate Gradient
(dx,dy) = np.gradient(grid_z1) #Find gradient for points on meshgrid
Velocity_dx= dx/stepx #velocity ms/m
Velocity_dy= dy/stepx #velocity ms/m
Resultant = (Velocity_dx**2 + Velocity_dy**2)**0.5 #Resultant scalar value ms/m
Resultant = np.ravel(Resultant)
#Plot Original Data F(X,Y) on the meshgrid
fig = pyplot.figure()
ax = fig.add_subplot(projection='3d')
ax.scatter(x,y,Arrival_Time,color='r')
ax.plot_trisurf(X, Y, Z)
ax.set_xlabel('X-Coordinates')
ax.set_ylabel('Y-Coordinates')
ax.set_zlabel('Time (ms)')
pyplot.show()
#Plot the Derivative of f'(X,Y) on the meshgrid
fig = pyplot.figure()
ax = fig.add_subplot(projection='3d')
ax.scatter(X,Y,Resultant,color='r',s=0.2)
ax.plot_trisurf(X, Y, Resultant)
ax.set_xlabel('X-Coordinates')
ax.set_ylabel('Y-Coordinates')
ax.set_zlabel('Velocity (ms/m)')
pyplot.show()
#Integrate to compare the original data input
dxintegral = np.nancumsum(Velocity_dx, axis=1)*stepx
dyintegral = np.nancumsum(Velocity_dy, axis=0)*stepy
valintegral = np.ma.zeros(dxintegral.shape)
for i in range(len(yy)):
for j in range(len(xx)):
valintegral[i, j] = np.ma.sum([dxintegral[0, len(xx) // 2],
dyintegral[i, len(yy) // 2], dxintegral[i, j], - dxintegral[i, len(xx) // 2]])
valintegral = valintegral * np.isfinite(dxintegral)
Now the np.gradient is applied at every meshnode (dx,dy) = np.gradient(grid_z1)
Now in my process I would analyse the gradient values above and make some adjustments (There is some unsual edge effects that are being create which I need to rectify) and would then integrate the values to get back to a surface which would be very similar to f(x,y) shown above.
I need some help adjusting the integration function:
#Integrate to compare the original data input
dxintegral = np.nancumsum(Velocity_dx, axis=1)*stepx
dyintegral = np.nancumsum(Velocity_dy, axis=0)*stepy
valintegral = np.ma.zeros(dxintegral.shape)
for i in range(len(yy)):
for j in range(len(xx)):
valintegral[i, j] = np.ma.sum([dxintegral[0, len(xx) // 2],
dyintegral[i, len(yy) // 2], dxintegral[i, j], - dxintegral[i, len(xx) // 2]])
valintegral = valintegral * np.isfinite(dxintegral)
And now I need to calculate the new 'Time' values at the original (x,y) point locations.
UPDATE (08-09-20) : I am getting some promising results using the help from #Aguy. The results can be seen below (with the blue contours representing the original data, and the red contours representing the integrated values).
I am still working on an integration approach which can remove the inaccuarcies at the areas of min(y) and max(y)
from matplotlib.tri import (Triangulation, UniformTriRefiner,
CubicTriInterpolator,LinearTriInterpolator,TriInterpolator,TriAnalyzer)
import pandas as pd
from scipy.interpolate import griddata
import matplotlib.pyplot as plt
import numpy as np
from scipy import interpolate
#-------------------------------------------------------------------------
# STEP 1: Import data from Excel file, and set variables
#-------------------------------------------------------------------------
df_initial = pd.read_excel(
r'C:\Users\morga\PycharmProjects\venv\Development\Trial'
r'.xlsx')
Inputdata can be found here link
df_initial = df_initial .sort_values(by='Delay', ascending=True) #Update dataframe and sort by Delay
x = df_initial ['X'].to_numpy()
y = df_initial ['Y'].to_numpy()
Arrival_Time = df_initial ['Delay'].to_numpy()
# Createmesh grid with a spacing of 0.5 x 0.5
stepx = 0.5
stepy = 0.5
xx = np.arange(min(x), max(x), stepx)
yy = np.arange(min(y), max(y), stepy)
xgrid, ygrid = np.meshgrid(xx, yy)
grid_z1 = interpolate.griddata((x, y), Arrival_Time, (xgrid, ygrid), method='linear') # Interpolating the Time values
# Calculate Gradient (velocity ms/m)
(dy, dx) = np.gradient(grid_z1) # Find gradient for points on meshgrid
Velocity_dx = dx / stepx # x velocity component ms/m
Velocity_dy = dy / stepx # y velocity component ms/m
# Integrate to compare the original data input
dxintegral = np.nancumsum(Velocity_dx, axis=1) * stepx
dyintegral = np.nancumsum(Velocity_dy, axis=0) * stepy
valintegral = np.ma.zeros(dxintegral.shape) # Makes an array filled with 0's the same shape as dx integral
for i in range(len(yy)):
for j in range(len(xx)):
valintegral[i, j] = np.ma.sum(
[dxintegral[0, len(xx) // 2], dyintegral[i, len(xx) // 2], dxintegral[i, j], - dxintegral[i, len(xx) // 2]])
valintegral[np.isnan(dx)] = np.nan
min_value = np.nanmin(valintegral)
valintegral = valintegral + (min_value * -1)
##Plot Results
fig = plt.figure()
ax = fig.add_subplot()
ax.scatter(x, y, color='black', s=7, zorder=3)
ax.set_xlabel('X-Coordinates')
ax.set_ylabel('Y-Coordinates')
ax.contour(xgrid, ygrid, valintegral, levels=50, colors='red', zorder=2)
ax.contour(xgrid, ygrid, grid_z1, levels=50, colors='blue', zorder=1)
ax.set_aspect('equal')
plt.show()
TL;DR;
You have multiple challenges to address in this issue, mainly:
Potential reconstruction (scalar field) from its gradient (vector field)
But also:
Observation in a concave hull with non rectangular grid;
Numerical 2D line integration and numerical inaccuracy;
It seems it can be solved by choosing an adhoc interpolant and a smart way to integrate (as pointed out by #Aguy).
MCVE
In a first time, let's build a MCVE to highlight above mentioned key points.
Dataset
We recreate a scalar field and its gradient.
import numpy as np
from scipy import interpolate
import matplotlib.pyplot as plt
def f(x, y):
return x**2 + x*y + 2*y + 1
Nx, Ny = 21, 17
xl = np.linspace(-3, 3, Nx)
yl = np.linspace(-2, 2, Ny)
X, Y = np.meshgrid(xl, yl)
Z = f(X, Y)
zl = np.arange(np.floor(Z.min()), np.ceil(Z.max())+1, 2)
dZdy, dZdx = np.gradient(Z, yl, xl, edge_order=1)
V = np.hypot(dZdx, dZdy)
The scalar field looks like:
axe = plt.axes(projection='3d')
axe.plot_surface(X, Y, Z, cmap='jet', alpha=0.5)
axe.view_init(elev=25, azim=-45)
And, the vector field looks like:
axe = plt.contour(X, Y, Z, zl, cmap='jet')
axe.axes.quiver(X, Y, dZdx, dZdy, V, units='x', pivot='tip', cmap='jet')
axe.axes.set_aspect('equal')
axe.axes.grid()
Indeed gradient is normal to potential levels. We also plot the gradient magnitude:
axe = plt.contour(X, Y, V, 10, cmap='jet')
axe.axes.set_aspect('equal')
axe.axes.grid()
Raw field reconstruction
If we naively reconstruct the scalar field from the gradient:
SdZx = np.cumsum(dZdx, axis=1)*np.diff(xl)[0]
SdZy = np.cumsum(dZdy, axis=0)*np.diff(yl)[0]
Zhat = np.zeros(SdZx.shape)
for i in range(Zhat.shape[0]):
for j in range(Zhat.shape[1]):
Zhat[i,j] += np.sum([SdZy[i,0], -SdZy[0,0], SdZx[i,j], -SdZx[i,0]])
Zhat += Z[0,0] - Zhat[0,0]
We can see the global result is roughly correct, but levels are less accurate where the gradient magnitude is low:
Interpolated field reconstruction
If we increase the grid resolution and pick a specific interpolant (usual when dealing with mesh grid), we can get a finer field reconstruction:
r = np.stack([X.ravel(), Y.ravel()]).T
Sx = interpolate.CloughTocher2DInterpolator(r, dZdx.ravel())
Sy = interpolate.CloughTocher2DInterpolator(r, dZdy.ravel())
Nx, Ny = 200, 200
xli = np.linspace(xl.min(), xl.max(), Nx)
yli = np.linspace(yl.min(), yl.max(), Nx)
Xi, Yi = np.meshgrid(xli, yli)
ri = np.stack([Xi.ravel(), Yi.ravel()]).T
dZdxi = Sx(ri).reshape(Xi.shape)
dZdyi = Sy(ri).reshape(Xi.shape)
SdZxi = np.cumsum(dZdxi, axis=1)*np.diff(xli)[0]
SdZyi = np.cumsum(dZdyi, axis=0)*np.diff(yli)[0]
Zhati = np.zeros(SdZxi.shape)
for i in range(Zhati.shape[0]):
for j in range(Zhati.shape[1]):
Zhati[i,j] += np.sum([SdZyi[i,0], -SdZyi[0,0], SdZxi[i,j], -SdZxi[i,0]])
Zhati += Z[0,0] - Zhati[0,0]
Which definitely performs way better:
So basically, increasing the grid resolution with an adhoc interpolant may help you to get more accurate result. The interpolant also solve the need to get a regular rectangular grid from a triangular mesh to perform integration.
Concave and convex hull
You also have pointed out inaccuracy on the edges. Those are the result of the combination of the interpolant choice and the integration methodology. The integration methodology fails to properly compute the scalar field when it reach concave region with few interpolated points. The problem disappear when choosing a mesh-free interpolant able to extrapolate.
To illustrate it, let's remove some data from our MCVE:
q = np.full(dZdx.shape, False)
q[0:6,5:11] = True
q[-6:,-6:] = True
dZdx[q] = np.nan
dZdy[q] = np.nan
Then the interpolant can be constructed as follow:
q2 = ~np.isnan(dZdx.ravel())
r = np.stack([X.ravel(), Y.ravel()]).T[q2,:]
Sx = interpolate.CloughTocher2DInterpolator(r, dZdx.ravel()[q2])
Sy = interpolate.CloughTocher2DInterpolator(r, dZdy.ravel()[q2])
Performing the integration we see that in addition of classical edge effect we do have less accurate value in concave regions (swingy dot-dash lines where the hull is concave) and we have no data outside the convex hull as Clough Tocher is a mesh-based interpolant:
Vl = np.arange(0, 11, 1)
axe = plt.contour(X, Y, np.hypot(dZdx, dZdy), Vl, cmap='jet')
axe.axes.contour(Xi, Yi, np.hypot(dZdxi, dZdyi), Vl, cmap='jet', linestyles='-.')
axe.axes.set_aspect('equal')
axe.axes.grid()
So basically the error we are seeing on the corner are most likely due to integration issue combined with interpolation limited to the convex hull.
To overcome this we can choose a different interpolant such as RBF (Radial Basis Function Kernel) which is able to create data outside the convex hull:
Sx = interpolate.Rbf(r[:,0], r[:,1], dZdx.ravel()[q2], function='thin_plate')
Sy = interpolate.Rbf(r[:,0], r[:,1], dZdy.ravel()[q2], function='thin_plate')
dZdxi = Sx(ri[:,0], ri[:,1]).reshape(Xi.shape)
dZdyi = Sy(ri[:,0], ri[:,1]).reshape(Xi.shape)
Notice the slightly different interface of this interpolator (mind how parmaters are passed).
The result is the following:
We can see the region outside the convex hull can be extrapolated (RBF are mesh free). So choosing the adhoc interpolant is definitely a key point to solve your problem. But we still need to be aware that extrapolation may perform well but is somehow meaningless and dangerous.
Solving your problem
The answer provided by #Aguy is perfectly fine as it setups a clever way to integrate that is not disturbed by missing points outside the convex hull. But as you mentioned there is inaccuracy in concave region inside the convex hull.
If you wish to remove the edge effect you detected, you will have to resort to an interpolant able to extrapolate as well, or find another way to integrate.
Interpolant change
Using RBF interpolant seems to solve your problem. Here is the complete code:
df = pd.read_excel('./Trial-Wireup 2.xlsx')
x = df['X'].to_numpy()
y = df['Y'].to_numpy()
z = df['Delay'].to_numpy()
r = np.stack([x, y]).T
#S = interpolate.CloughTocher2DInterpolator(r, z)
#S = interpolate.LinearNDInterpolator(r, z)
S = interpolate.Rbf(x, y, z, epsilon=0.1, function='thin_plate')
N = 200
xl = np.linspace(x.min(), x.max(), N)
yl = np.linspace(y.min(), y.max(), N)
X, Y = np.meshgrid(xl, yl)
#Zp = S(np.stack([X.ravel(), Y.ravel()]).T)
Zp = S(X.ravel(), Y.ravel())
Z = Zp.reshape(X.shape)
dZdy, dZdx = np.gradient(Z, yl, xl, edge_order=1)
SdZx = np.nancumsum(dZdx, axis=1)*np.diff(xl)[0]
SdZy = np.nancumsum(dZdy, axis=0)*np.diff(yl)[0]
Zhat = np.zeros(SdZx.shape)
for i in range(Zhat.shape[0]):
for j in range(Zhat.shape[1]):
#Zhat[i,j] += np.nansum([SdZy[i,0], -SdZy[0,0], SdZx[i,j], -SdZx[i,0]])
Zhat[i,j] += np.nansum([SdZx[0,N//2], SdZy[i,N//2], SdZx[i,j], -SdZx[i,N//2]])
Zhat += Z[100,100] - Zhat[100,100]
lz = np.linspace(0, 5000, 20)
axe = plt.contour(X, Y, Z, lz, cmap='jet')
axe = plt.contour(X, Y, Zhat, lz, cmap='jet', linestyles=':')
axe.axes.plot(x, y, '.', markersize=1)
axe.axes.set_aspect('equal')
axe.axes.grid()
Which graphically renders as follow:
The edge effect is gone because of the RBF interpolant can extrapolate over the whole grid. You can confirm it by comparing the result of mesh-based interpolants.
Linear
Clough Tocher
Integration variable order change
We can also try to find a better way to integrate and mitigate the edge effect, eg. let's change the integration variable order:
Zhat[i,j] += np.nansum([SdZy[N//2,0], SdZx[N//2,j], SdZy[i,j], -SdZy[N//2,j]])
With a classic linear interpolant. The result is quite correct, but we still have an edge effect on the bottom left corner:
As you noticed the problem occurs at the middle of the axis in region where the integration starts and lacks a reference point.
Here is one approach:
First, in order to be able to do integration, it's good to be on a regular grid. Using here variable names x and y as short for your triang.x and triang.y we can first create a grid:
import numpy as np
n = 200 # Grid density
stepx = (max(x) - min(x)) / n
stepy = (max(y) - min(y)) / n
xspace = np.arange(min(x), max(x), stepx)
yspace = np.arange(min(y), max(y), stepy)
xgrid, ygrid = np.meshgrid(xspace, yspace)
Then we can interpolate dx and dy on the grid using the same LinearTriInterpolator function:
fdx = LinearTriInterpolator(masked_triang, dx)
fdy = LinearTriInterpolator(masked_triang, dy)
dxgrid = fdx(xgrid, ygrid)
dygrid = fdy(xgrid, ygrid)
Now comes the integration part. In principle, any path we choose should get us to the same value. In practice, since there are missing values and different densities, the choice of path is very important to get a reasonably accurate answer.
Below I choose to integrate over dxgrid in the x direction from 0 to the middle of the grid at n/2. Then integrate over dygrid in the y direction from 0 to the i point of interest. Then over dxgrid again from n/2 to the point j of interest. This is a simple way to make sure most of the path of integration is inside the bulk of available data by simply picking a path that goes mostly in the "middle" of the data range. Other alternative consideration would lead to different path selections.
So we do:
dxintegral = np.nancumsum(dxgrid, axis=1) * stepx
dyintegral = np.nancumsum(dygrid, axis=0) * stepy
and then (by somewhat brute force for clarity):
valintegral = np.ma.zeros(dxintegral.shape)
for i in range(n):
for j in range(n):
valintegral[i, j] = np.ma.sum([dxintegral[0, n // 2], dyintegral[i, n // 2], dxintegral[i, j], - dxintegral[i, n // 2]])
valintegral = valintegral * np.isfinite(dxintegral)
valintegral would be the result up to an arbitrary constant which can help put the "zero" where you want.
With your data shown here:
ax.tricontourf(masked_triang, time_array)
This is what I'm getting reconstructed when using this method:
ax.contourf(xgrid, ygrid, valintegral)
Hopefully this is somewhat helpful.
If you want to revisit the values at the original triangulation points, you can use interp2d on the valintegral regular grid data.
EDIT:
In reply to your edit, your adaptation above has a few errors:
Change the line (dx,dy) = np.gradient(grid_z1) to (dy,dx) = np.gradient(grid_z1)
In the integration loop change the dyintegral[i, len(yy) // 2] term to dyintegral[i, len(xx) // 2]
Better to replace the line valintegral = valintegral * np.isfinite(dxintegral) with valintegral[np.isnan(dx)] = np.nan
I am trying to find a python package that would give an option to fit natural smoothing splines with user selectable smoothing factor. Is there an implementation for that? If not, how would you use what is available to implement it yourself?
By natural spline I mean that there should be a condition that the second derivative of the fitted function at the endpoints is zero (linear).
By smoothing spline I mean that the spline should not be 'interpolating' (passing through all the datapoints). I would like to decide the correct smoothing factor lambda (see the Wikipedia page for smoothing splines) myself.
What I have found
scipy.interpolate.CubicSpline [link]: Does natural (cubic) spline fitting. Does interpolation, and there is no way to smooth the data.
scipy.interpolate.UnivariateSpline [link]: Does spline fitting with user selectable smoothing factor. However, there is no option to make the splines natural.
After hours of investigation, I did not find any pip installable packages which could fit a natural cubic spline with user-controllable smoothness. However, after deciding to write one myself, while reading about the topic I stumbled upon a blog post by github user madrury. He has written python code capable of producing natural cubic spline models.
The model code is available here (NaturalCubicSpline) with a BSD-licence. He has also written some examples in an IPython notebook.
But since this is the Internet and links tend to die, I will copy the relevant parts of the source code here + a helper function (get_natural_cubic_spline_model) written by me, and show an example of how to use it. The smoothness of the fit can be controlled by using different number of knots. The position of the knots can be also specified by the user.
Example
from matplotlib import pyplot as plt
import numpy as np
def func(x):
return 1/(1+25*x**2)
# make example data
x = np.linspace(-1,1,300)
y = func(x) + np.random.normal(0, 0.2, len(x))
# The number of knots can be used to control the amount of smoothness
model_6 = get_natural_cubic_spline_model(x, y, minval=min(x), maxval=max(x), n_knots=6)
model_15 = get_natural_cubic_spline_model(x, y, minval=min(x), maxval=max(x), n_knots=15)
y_est_6 = model_6.predict(x)
y_est_15 = model_15.predict(x)
plt.plot(x, y, ls='', marker='.', label='originals')
plt.plot(x, y_est_6, marker='.', label='n_knots = 6')
plt.plot(x, y_est_15, marker='.', label='n_knots = 15')
plt.legend(); plt.show()
The source code for get_natural_cubic_spline_model
import numpy as np
import pandas as pd
from sklearn.base import BaseEstimator, TransformerMixin
from sklearn.linear_model import LinearRegression
from sklearn.pipeline import Pipeline
def get_natural_cubic_spline_model(x, y, minval=None, maxval=None, n_knots=None, knots=None):
"""
Get a natural cubic spline model for the data.
For the knots, give (a) `knots` (as an array) or (b) minval, maxval and n_knots.
If the knots are not directly specified, the resulting knots are equally
space within the *interior* of (max, min). That is, the endpoints are
*not* included as knots.
Parameters
----------
x: np.array of float
The input data
y: np.array of float
The outpur data
minval: float
Minimum of interval containing the knots.
maxval: float
Maximum of the interval containing the knots.
n_knots: positive integer
The number of knots to create.
knots: array or list of floats
The knots.
Returns
--------
model: a model object
The returned model will have following method:
- predict(x):
x is a numpy array. This will return the predicted y-values.
"""
if knots:
spline = NaturalCubicSpline(knots=knots)
else:
spline = NaturalCubicSpline(max=maxval, min=minval, n_knots=n_knots)
p = Pipeline([
('nat_cubic', spline),
('regression', LinearRegression(fit_intercept=True))
])
p.fit(x, y)
return p
class AbstractSpline(BaseEstimator, TransformerMixin):
"""Base class for all spline basis expansions."""
def __init__(self, max=None, min=None, n_knots=None, n_params=None, knots=None):
if knots is None:
if not n_knots:
n_knots = self._compute_n_knots(n_params)
knots = np.linspace(min, max, num=(n_knots + 2))[1:-1]
max, min = np.max(knots), np.min(knots)
self.knots = np.asarray(knots)
#property
def n_knots(self):
return len(self.knots)
def fit(self, *args, **kwargs):
return self
class NaturalCubicSpline(AbstractSpline):
"""Apply a natural cubic basis expansion to an array.
The features created with this basis expansion can be used to fit a
piecewise cubic function under the constraint that the fitted curve is
linear *outside* the range of the knots.. The fitted curve is continuously
differentiable to the second order at all of the knots.
This transformer can be created in two ways:
- By specifying the maximum, minimum, and number of knots.
- By specifying the cutpoints directly.
If the knots are not directly specified, the resulting knots are equally
space within the *interior* of (max, min). That is, the endpoints are
*not* included as knots.
Parameters
----------
min: float
Minimum of interval containing the knots.
max: float
Maximum of the interval containing the knots.
n_knots: positive integer
The number of knots to create.
knots: array or list of floats
The knots.
"""
def _compute_n_knots(self, n_params):
return n_params
#property
def n_params(self):
return self.n_knots - 1
def transform(self, X, **transform_params):
X_spl = self._transform_array(X)
if isinstance(X, pd.Series):
col_names = self._make_names(X)
X_spl = pd.DataFrame(X_spl, columns=col_names, index=X.index)
return X_spl
def _make_names(self, X):
first_name = "{}_spline_linear".format(X.name)
rest_names = ["{}_spline_{}".format(X.name, idx)
for idx in range(self.n_knots - 2)]
return [first_name] + rest_names
def _transform_array(self, X, **transform_params):
X = X.squeeze()
try:
X_spl = np.zeros((X.shape[0], self.n_knots - 1))
except IndexError: # For arrays with only one element
X_spl = np.zeros((1, self.n_knots - 1))
X_spl[:, 0] = X.squeeze()
def d(knot_idx, x):
def ppart(t): return np.maximum(0, t)
def cube(t): return t*t*t
numerator = (cube(ppart(x - self.knots[knot_idx]))
- cube(ppart(x - self.knots[self.n_knots - 1])))
denominator = self.knots[self.n_knots - 1] - self.knots[knot_idx]
return numerator / denominator
for i in range(0, self.n_knots - 2):
X_spl[:, i+1] = (d(i, X) - d(self.n_knots - 2, X)).squeeze()
return X_spl
You could use this numpy/scipy implementation of natural cubic smoothing spline for univariate/multivariate data smoothing. Smoothing parameter should be in range [0.0, 1.0]. If we use smoothing parameter equal to 1.0 we get natural cubic spline interpolant without data smoothing. Also the implementation supports vectorization for univariate data.
Univariate example:
import numpy as np
import matplotlib.pyplot as plt
import csaps
np.random.seed(1234)
x = np.linspace(-5., 5., 25)
y = np.exp(-(x/2.5)**2) + (np.random.rand(25) - 0.2) * 0.3
sp = csaps.UnivariateCubicSmoothingSpline(x, y, smooth=0.85)
xs = np.linspace(x[0], x[-1], 150)
ys = sp(xs)
plt.plot(x, y, 'o', xs, ys, '-')
plt.show()
Bivariate example:
import numpy as np
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
import csaps
xdata = [np.linspace(-3, 3, 61), np.linspace(-3.5, 3.5, 51)]
i, j = np.meshgrid(*xdata, indexing='ij')
ydata = (3 * (1 - j)**2. * np.exp(-(j**2) - (i + 1)**2)
- 10 * (j / 5 - j**3 - i**5) * np.exp(-j**2 - i**2)
- 1 / 3 * np.exp(-(j + 1)**2 - i**2))
np.random.seed(12345)
noisy = ydata + (np.random.randn(*ydata.shape) * 0.75)
sp = csaps.MultivariateCubicSmoothingSpline(xdata, noisy, smooth=0.988)
ysmth = sp(xdata)
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
ax.plot_wireframe(j, i, noisy, linewidths=0.5, color='r')
ax.scatter(j, i, noisy, s=5, c='r')
ax.plot_surface(j, i, ysmth, linewidth=0, alpha=1.0)
plt.show()
The python package patsy has functions for generating spline bases, including a natural cubic spline basis. Described in the documentation.
Any library can then be used for fitting a model, e.g. scikit-learn or statsmodels.
The df parameter for cr() can be used to control the "smoothness"
Note that too low df can result to underfit (see below).
A simple example using scikit-learn.
import numpy as np
from sklearn.linear_model import LinearRegression
from patsy import cr
import matplotlib.pyplot as plt
n_obs = 600
np.random.seed(0)
x = np.linspace(-3, 3, n_obs)
y = 1 / (x ** 2 + 1) * np.cos(np.pi * x) + np.random.normal(0, 0.2, size=n_obs)
def plot_smoothed(df=5):
# Generate spline basis with different degrees of freedom
x_basis = cr(x, df=df, constraints="center")
# Fit model to the data
model = LinearRegression().fit(x_basis, y)
# Get estimates
y_hat = model.predict(x_basis)
plt.plot(x, y_hat, label=f"df={df}")
plt.scatter(x, y, s=4, color="tab:blue")
for df in (5, 7, 10, 25):
plot_smoothed(df)
plt.legend()
plt.title(f"Natural cubic spline with varying degrees of freedom")
plt.show()
For a project of mine, I needed to create intervals for time-series modeling, and to make the procedure more efficient I created tsmoothie: A python library for time-series smoothing and outlier detection in a vectorized way.
It provides different smoothing algorithms together with the possibility to computes intervals.
In the case of SplineSmoother of natural cubic type:
import numpy as np
import matplotlib.pyplot as plt
from tsmoothie.smoother import *
def func(x):
return 1/(1+25*x**2)
# make example data
x = np.linspace(-1,1,300)
y = func(x) + np.random.normal(0, 0.2, len(x))
# operate smoothing
smoother = SplineSmoother(n_knots=10, spline_type='natural_cubic_spline')
smoother.smooth(y)
# generate intervals
low, up = smoother.get_intervals('prediction_interval', confidence=0.05)
# plot the first smoothed timeseries with intervals
plt.figure(figsize=(11,6))
plt.plot(smoother.smooth_data[0], linewidth=3, color='blue')
plt.plot(smoother.data[0], '.k')
plt.fill_between(range(len(smoother.data[0])), low[0], up[0], alpha=0.3)
I point out also that tsmoothie can carry out the smoothing of multiple time-series in a vectorized way
The programming language R offers a very good implementation of natural cubic smoothing splines. You can use R functions in Python with rpy2:
import rpy2.robjects as robjects
r_y = robjects.FloatVector(y_train)
r_x = robjects.FloatVector(x_train)
r_smooth_spline = robjects.r['smooth.spline'] #extract R function# run smoothing function
spline1 = r_smooth_spline(x=r_x, y=r_y, spar=0.7)
ySpline=np.array(robjects.r['predict'](spline1,robjects.FloatVector(x_smooth)).rx2('y'))
plt.plot(x_smooth,ySpline)
If you want to directly set lambda: spline1 = r_smooth_spline(x=r_x, y=r_y, lambda=42) doesn't work, because lambda has already another meaning in Python, but there is a solution: How to use the lambda argument of smooth.spline in RPy WITHOUT Python interprating it as lambda.
To get the code running you first need to define the data x_train and y_train and you can define x_smooth=np.array(np.linspace(-3,5,1920)). if you want to plot it between -3 and 5 in Full-HD-resolution.
Note that this code is not fully compatible with Jupyter-notebooks for the latest versions of rpy2. You can fix this by using !pip install -Iv rpy2==3.4.2 as described in NotImplementedError: Conversion 'rpy2py' not defined for objects of type '<class 'rpy2.rinterface.SexpClosure'>' only after I run the code twice
I want to generate a Gaussian distribution in Python with the x and y dimensions denoting position and the z dimension denoting the magnitude of a certain quantity.
The distribution has a maximum value of 2e6 and a standard deviation sigma=0.025.
In MATLAB I can do this with:
x1 = linspace(-1,1,30);
x2 = linspace(-1,1,30);
mu = [0,0];
Sigma = [.025,.025];
[X1,X2] = meshgrid(x1,x2);
F = mvnpdf([X1(:) X2(:)],mu,Sigma);
F = 314159.153*reshape(F,length(x2),length(x1));
surf(x1,x2,F);
In Python, what I have so far is:
x = np.linspace(-1,1,30)
y = np.linspace(-1,1,30)
mu = (np.median(x),np.median(y))
sigma = (.025,.025)
There is a Numpy function numpy.random.multivariate_normal what can supposedly do the same as MATLAB's mvnpdf, but I am struggling to undestand the documentation. Especially in obtaining the covariance matrix needed by numpy.random.multivariate_normal.
As of scipy 0.14, you can use scipy.stats.multivariate_normal.pdf()
import numpy as np
from scipy.stats import multivariate_normal
x, y = np.mgrid[-1.0:1.0:30j, -1.0:1.0:30j]
# Need an (N, 2) array of (x, y) pairs.
xy = np.column_stack([x.flat, y.flat])
mu = np.array([0.0, 0.0])
sigma = np.array([.025, .025])
covariance = np.diag(sigma**2)
z = multivariate_normal.pdf(xy, mean=mu, cov=covariance)
# Reshape back to a (30, 30) grid.
z = z.reshape(x.shape)
I am working on a scikit called scikit-guess that contains some fast estimation routines for non-linear fits. It has a function skg.ngauss.model (also accessible as skg.ngauss_fit.model or skg.ngauss.ngauss_fit.model) which does exactly what you want. The nice thing is that it's not a PDF, so you set the amplitude out of the box:
import numpy as np
import skg.ngauss
a = 2e6
mu = 0, 0
sigma = 0.025, 0.025
x = y = np.linspace(-1, 1, 31)
cov = np.diag(sigma)**2
X = np.meshgrid(x, y)
data = skg.ngauss.model(X, a, mu, cov, axis=0)
You need to tell it axis=0 because it automatically stacks your arrays for you. To avoid passing in that argument, you could write
X = np.stack(np.meshgrid(x, y), axis=-1)
You can plot the result:
from matplotlib import pyplot as plt
plt.imshow(data)
plt.show()
This is not a very exciting distribution because the spread is so small that you end up with a value of ~2e-5 just one pixel away. You may want to up your sampling space to get any sort of meaningful resolution.
Note: At time of writing, the fitting function (ngauss_fit) is still buggy, but the model has been tested successfully, just not in the scikit.
Disclaimer: In case it wasn't obvious from the above, I am the author of scikit-guess.