Develop Reference

Double antiderivative computation in python

I have the following problem. I have a function f defined in python using numpy functions. The function is smooth and integrable on positive reals. I want to construct the double antiderivative of the function (assuming that both the value and the slope of the antiderivative at 0 are 0) so that I can evaluate it on any positive real smaller than 100.
Definition of antiderivative of f at x:
integrate f(s) with s from 0 to x
Definition of double antiderivative of f at x:
integrate (integrate f(t) with t from 0 to s) with s from 0 to x
The actual form of f is not important, so I will use a simple one for convenience. But please note that even though my example has a known closed form, my actual function does not.
import numpy as np
f = lambda x: np.exp(-x)*x
My solution is to construct the antiderivative as an array using naive numerical integration:
N = 10000
delta = 100/N
xs = np.linspace(0,100,N+1)
vs = f(xs)
avs = np.cumsum(vs)*delta
aavs = np.cumsum(avs)*delta
This of course works but it gives me arrays instead of functions. But this is not a big problem as I can interpolate aavs using a spline to get a function and get rid of the arrays.
from scipy.interpolate import UnivariateSpline
aaf = UnivariateSpline(xs, aavs)
The function aaf is approximately the double antiderivative of f.
The problem is that even though it works, there is quite a bit of overhead before I can get my function and precision is expensive.
My other idea was to interpolate f by a spline and take the antiderivative of that, however this introduces numerical errors that are too big for what I want to use the function.
Is there any better way to do that? By better I mean faster without sacrificing accuracy.
Edit: What I hope is possible is to use some kind of Fourier transform to avoid integrating twice. I hope that there is some convenient transform of vs that allows to multiply the values component-wise with xs and transform back to get the double antiderivative. I played with this a bit, but I got lost.
Edit: I figured out that by using the trapezoidal rule instead of a naive sum, increases the accuracy quite a bit. Using Simpson's rule should increase the accuracy further, but it's somewhat fiddly to do with numpy arrays.
Edit: As #user202729 rightfully complains, this seems off. The reason it seems off is because I have skipped some details. I explain here why what I say makes sense, but it does not affect my question.
My actual goal is not to find the double antiderivative of f, but to find a transformation of this. I have skipped that because I think it only confuses the matter.
The function f decays exponentially as x approaches 0 or infinity. I am minimizing the numerical error in the integration by starting the sum from 0 and going up to approximately the peak of f. This ensure that the relative error is approximately constant. Then I start from the opposite direction from some very big x and go back to the peak. Then I do the same for the antiderivative values.
Then I transform the aavs by another function which is sensitive to numerical errors. Then I find the region where the errors are big (the values oscillate violently) and drop these values. Finally I approximate what I believe are good values by a spline.
Now if I use spline to approximate f, it introduces an absolute error which is the dominant term in a rather large interval. This gets "integrated" twice and it ends up being a rather large relative error in aavs. Then once I transform aavs, I find that the 'good region' has shrunk considerably.
EDIT: The actual form of f is something I'm still looking into. However, it is going to be a generalisation of the lognormal distribution. Right now I am playing with the following family.
I start by defining a generalization of the normal distribution:
def pdf_n(params, center=0.0, slope=8):
scale, min, diff = params
if diff > 0:
r = min
l = min + diff
else:
r = min - diff
l = min
def retfun(m):
x = (m - center)/scale
E = special.expit(slope*x)*(r - l) + l
return np.exp( -np.power(1 + x*x, E)/2 )
return np.vectorize(retfun)
It may not be obvious what is happening here, but the result is quite simple. The function decays as exp(-x^(2l)) on the left and as exp(-x^(2r)) on the right. For min=1 and diff=0, this is the normal distribution. Note that this is not normalized. Then I define
g = pdf(params)
f = np.vectorize(lambda x:g(np.log(x))/x/area)
where area is the normalization constant.
Note that this is not the actual code I use. I stripped it down to the bare minimum.

You can compute the two np.cumsum (and the divisions) at once more efficiently using Numba. This is significantly faster since there is no need for several temporary arrays to be allocated, filled, read again and freed. Here is a naive implementation:
import numba as nb
#nb.njit('float64[::1](float64[::1], float64)') # Assume vs is contiguous
def doubleAntiderivative_naive(vs, delta):
res = np.empty(vs.size, dtype=np.float64)
sum1, sum2 = 0.0, 0.0
for i in range(vs.size):
sum1 += vs[i] * delta
sum2 += sum1 * delta
res[i] = sum2
return res
However, the sum is not very good in term of numerical stability. A Kahan summation is needed to improve the accuracy (or possibly the alternative Kahan–Babuška-Klein algorithm if you are paranoid about the accuracy and performance do not matter so much). Note that Numpy use a pair-wise algorithm which is quite good but far from being prefect in term of accuracy (this is a good compromise for both performance and accuracy).
Moreover, delta can be factorized during in the summation (ie. the result just need to be premultiplied by delta**2).
Here is an implementation using the more accurate Kahan summation:
#nb.njit('float64[::1](float64[::1], float64)')
def doubleAntiderivative_accurate(vs, delta):
res = np.empty(vs.size, dtype=np.float64)
delta2 = delta * delta
sum1, sum2 = 0.0, 0.0
c1, c2 = 0.0, 0.0
for i in range(vs.size):
# Kahan summation of the antiderivative of vs
y1 = vs[i] - c1
t1 = sum1 + y1
c1 = (t1 - sum1) - y1
sum1 = t1
# Kahan summation of the double antiderivative of vs
y2 = sum1 - c2
t2 = sum2 + y2
c2 = (t2 - sum2) - y2
sum2 = t2
res[i] = sum2 * delta2
return res
Here is the performance of the approaches on my machine (with an i5-9600KF processor):
Numpy cumsum: 51.3 us
Naive Numba: 11.6 us
Accutate Numba: 37.2 us
Here is the relative error of the approaches (based on the provided input function):
Numpy cumsum: 1e-13
Naive Numba: 5e-14
Accutate Numba: 2e-16
Perfect precision: 1e-16 (assuming 64-bit numbers are used)
If f can be easily computed using Numba (this is the case here), then vs[i] can be replaced by calls to f (inlined by Numba). This helps to reduce the memory consumption of the computation (N can be huge without saturating your RAM).
As for the interpolation, the splines often gives good numerical result but they are quite expensive to compute and AFAIK they require the whole array to be computed (each item of the array impact all the spline although some items may have a negligible impact alone). Regarding your needs, you could consider using Lagrange polynomials. You should be careful when using Lagrange polynomials on the edges. In your case, you can easily solve the numerical divergence issue on the edges by extending the array size with the border values (since you know the derivative on each edges of vs is 0). You can apply the interpolation on the fly with this method which can be good for both performance (typically if the computation is parallelized) and memory usage.

First, I created a version of the code I found more intuitive. Here I multiply cumulative sum values by bin widths. I believe there is a small error in the original version of the code related to the bin width issue.
import numpy as np
f = lambda x: np.exp(-x)*x
N = 1000
xs = np.linspace(0,100,N+1)
domainwidth = ( np.max(xs) - np.min(xs) )
binwidth = domainwidth / N
vs = f(xs)
avs = np.cumsum(vs)*binwidth
aavs = np.cumsum(avs)*binwidth
Next, for visualization here is some very simple plotting code:
import matplotlib
import matplotlib.pyplot as plt
plt.figure()
plt.scatter( xs, vs )
plt.figure()
plt.scatter( xs, avs )
plt.figure()
plt.scatter( xs, aavs )
plt.show()
The first integral matches the known result of the example expression and can be seen on wolfram
Below is a simple function that extracts an element from the second derivative. Note that int is a bad rounding function. I assume this is what you have implemented already.
def extract_double_antideriv_value(x):
return aavs[int(x/binwidth)]
singleresult = extract_double_antideriv_value(50.24)
print('singleresult', singleresult)
Whatever full computation steps are required, we need to know them before we can start optimizing. Do you have a million different functions to integrate? If you only need to query a single double anti-derivative many times, your original solution should be fairly ideal.
Symbolic Approximation:
Have you considered approximations to the original function f, which can have closed form integration solutions? You have a limited domain on which the function lives. Perhaps approximate f with a Taylor series (which can be constructed with known maximum error) then integrate exactly? (consider Pade, Taylor, Fourier, Cheby, Lagrange(as suggested by another answer), etc...)
Log Tricks:
Another alternative to dealing with spiky errors, would be to take the log of your original function. Is f always positive? Is the integration error caused because the neighborhood around the max is very small? If so, you can study ln(f) or even ln(ln(f)) instead. It would really help to understand what f looks like more.
Approximation Integration Tricks
There exist countless integration tricks in general, which can make approximate closed form solutions to undo-able integrals. A very common one when exponetnial functions are involved (I think yours is expoential?) is to use Laplace's Method. But which trick to pull out of the bag is highly dependent upon the conditions which f satisfies.

Performance issue with Scipy's solve_bvp and coupled differential equations

I'm facing a problem while trying to implement the coupled differential equation below (also known as single-mode coupling equation) in Python 3.8.3. As for the solver, I am using Scipy's function scipy.integrate.solve_bvp, whose documentation can be read here. I want to solve the equations in the complex domain, for different values of the propagation axis (z) and different values of beta (beta_analysis).
The problem is that it is extremely slow (not manageable) compared with an equivalent implementation in Matlab using the functions bvp4c, bvpinit and bvpset. Evaluating the first few iterations of both executions, they return the same result, except for the resulting mesh which is a lot greater in the case of Scipy. The mesh sometimes even saturates to the maximum value.
The equation to be solved is shown here below, along with the boundary conditions function.
import h5py
import numpy as np
from scipy import integrate
def coupling_equation(z_mesh, a):
ka_z = k # Global
z_a = z # Global
a_p = np.empty_like(a).astype(complex)
for idx, z_i in enumerate(z_mesh):
beta_zf_i = np.interp(z_i, z_a, beta_zf) # Get beta at the desired point of the mesh
ka_z_i = np.interp(z_i, z_a, ka_z) # Get ka at the desired point of the mesh
coupling_matrix = np.empty((2, 2), complex)
coupling_matrix[0] = [-1j * beta_zf_i, ka_z_i]
coupling_matrix[1] = [ka_z_i, 1j * beta_zf_i]
a_p[:, idx] = np.matmul(coupling_matrix, a[:, idx]) # Solve the coupling matrix
return a_p
def boundary_conditions(a_a, a_b):
return np.hstack(((a_a[0]-1), a_b[1]))
Moreover, I couldn't find a way to pass k, z and beta_zf as arguments of the function coupling_equation, given that the fun argument of the solve_bpv function must be a callable with the parameters (x, y). My approach is to define some global variables, but I would appreciate any help on this too if there is a better solution.
The analysis function which I am trying to code is:
def analysis(k, z, beta_analysis, max_mesh):
s11_analysis = np.empty_like(beta_analysis, dtype=complex)
s21_analysis = np.empty_like(beta_analysis, dtype=complex)
initial_mesh = np.linspace(z[0], z[-1], 10) # Initial mesh of 10 samples along L
mesh = initial_mesh
# a_init must be complex in order to solve the problem in a complex domain
a_init = np.vstack((np.ones(np.size(initial_mesh)).astype(complex),
np.zeros(np.size(initial_mesh)).astype(complex)))
for idx, beta in enumerate(beta_analysis):
print(f"Iteration {idx}: beta_analysis = {beta}")
global beta_zf
beta_zf = beta * np.ones(len(z)) # Global variable so as to use it in coupling_equation(x, y)
a = integrate.solve_bvp(fun=coupling_equation,
bc=boundary_conditions,
x=mesh,
y=a_init,
max_nodes=max_mesh,
verbose=1)
# mesh = a.x # Mesh for the next iteration
# a_init = a.y # Initial guess for the next iteration, corresponding to the current solution
s11_analysis[idx] = a.y[1][0]
s21_analysis[idx] = a.y[0][-1]
return s11_analysis, s21_analysis
I suspect that the problem has something to do with the initial guess that is being passed to the different iterations (see commented lines inside the loop in the analysis function). I try to set the solution of an iteration as the initial guess for the following (which must reduce the time needed for the solver), but it is even slower, which I don't understand. Maybe I missed something, because it is my first time trying to solve differential equations.
The parameters used for the execution are the following:
f2 = h5py.File(r'path/to/file', 'r')
k = np.array(f2['k']).squeeze()
z = np.array(f2['z']).squeeze()
f2.close()
analysis_points = 501
max_mesh = 1e6
beta_0 = 3e2;
beta_low = 0; # Lower value of the frequency for the analysis
beta_up = beta_0; # Upper value of the frequency for the analysis
beta_analysis = np.linspace(beta_low, beta_up, analysis_points);
s11_analysis, s21_analysis = analysis(k, z, beta_analysis, max_mesh)
Any ideas on how to improve the performance of these functions? Thank you all in advance, and sorry if the question is not well-formulated, I accept any suggestions about this.
Edit: Added some information about performance and sizing of the problem.
In practice, I can't find a relation that determines de number of times coupling_equation is called. It must be a matter of the internal operation of the solver. I checked the number of callings in one iteration by printing a line, and it happened in 133 ocasions (this was one of the fastests). This must be multiplied by the number of iterations of beta. For the analyzed one, the solver returned this:
Solved in 11 iterations, number of nodes 529.
Maximum relative residual: 9.99e-04
Maximum boundary residual: 0.00e+00
The shapes of a and z_mesh are correlated, since z_mesh is a vector whose length corresponds with the size of the mesh, recalculated by the solver each time it calls coupling_equation. Given that a contains the amplitudes of the progressive and regressive waves at each point of z_mesh, the shape of a is (2, len(z_mesh)).
In terms of computation times, I only managed to achieve 19 iterations in about 2 hours with Python. In this case, the initial iterations were faster, but they start to take more time as their mesh grows, until the point that the mesh saturates to the maximum allowed value. I think this is because of the value of the input coupling coefficients in that point, because it also happens when no loop in beta_analysisis executed (just the solve_bvp function for the intermediate value of beta). Instead, Matlab managed to return a solution for the entire problem in just 6 minutes, aproximately. If I pass the result of the last iteration as initial_guess (commented lines in the analysis function, the mesh overflows even faster and it is impossible to get more than a couple iterations.

Based on semi-random inputs, we can see that max_mesh is sometimes reached. This means that coupling_equation can be called with a quite big z_mesh and a arrays. The problem is that coupling_equation contains a slow pure-Python loop iterating on each column of the arrays. You can speed the computation up a lot using Numpy vectorization. Here is an implementation:
def coupling_equation_fast(z_mesh, a):
ka_z = k # Global
z_a = z # Global
a_p = np.empty(a.shape, dtype=np.complex128)
beta_zf_i = np.interp(z_mesh, z_a, beta_zf) # Get beta at the desired point of the mesh
ka_z_i = np.interp(z_mesh, z_a, ka_z) # Get ka at the desired point of the mesh
# Fast manual matrix multiplication
a_p[0] = (-1j * beta_zf_i) * a[0] + ka_z_i * a[1]
a_p[1] = ka_z_i * a[0] + (1j * beta_zf_i) * a[1]
return a_p
This code provides a similar output with semi-random inputs compared to the original implementation but is roughly 20 times faster on my machine.
Furthermore, I do not know if max_mesh happens to be big with your inputs too and even if this is normal/intended. It may make sense to decrease the value of max_mesh in order to reduce the execution time even more.

Why ifft2 is not working but fft2 is fine?

I want to implement ifft2 using DFT matrix. The following code works for fft2.
import numpy as np
def DFT_matrix(N):
i, j = np.meshgrid(np.arange(N), np.arange(N))
omega = np.exp( - 2 * np.pi * 1j / N )
W = np.power( omega, i * j ) # Normalization by sqrt(N) Not included
return W
sizeM=40
sizeN=20
np.random.seed(0)
rA=np.random.rand(sizeM,sizeN)
rAfft=np.fft.fft2(rA)
dftMtxM=DFT_matrix(sizeM)
dftMtxN=DFT_matrix(sizeN)
# Matrix multiply the 3 matrices together
mA = dftMtxM # rA # dftMtxN
print(np.allclose(np.abs(mA), np.abs(rAfft)))
print(np.allclose(np.angle(mA), np.angle(rAfft)))
To get to ifft2 I assumd I need to change only the dft matrix to it's transpose, so expected the following to work, but I got false for the last two print any suggesetion please?
import numpy as np
def DFT_matrix(N):
i, j = np.meshgrid(np.arange(N), np.arange(N))
omega = np.exp( - 2 * np.pi * 1j / N )
W = np.power( omega, i * j ) # Normalization by sqrt(N) Not included
return W
sizeM=40
sizeN=20
np.random.seed(0)
rA=np.random.rand(sizeM,sizeN)
rAfft=np.fft.ifft2(rA)
dftMtxM=np.conj(DFT_matrix(sizeM))
dftMtxN=np.conj(DFT_matrix(sizeN))
# Matrix multiply the 3 matrices together
mA = dftMtxM # rA # dftMtxN
print(np.allclose(np.abs(mA), np.abs(rAfft)))
print(np.allclose(np.angle(mA), np.angle(rAfft)))

I am going to be building on some things from my answer to your previous question. Please note that I will try to distinguish between the terms Discrete Fourier Transform (DFT) and Fast Fourier Transform (FFT). Remember that DFT is the transform while FFT is only an efficient algorithm for performing it. People, including myself, however very commonly refer to the DFT as FFT since it is practically the only algorithm used for computing the DFT
The problem here is again the normalization of the data. It's interesting that this is such a fundamental and confusing part of any DFT operations yet I couldn't find a good explanation on the internet. I will try to provide a summary at the end about DFT normalization however I think the best way to understand this is by working through some examples yourself.
Why the comparisons fail?
It's important to note, that even though both of the allclose tests seemingly fail, they are actually not a very good method of comparing two complex number arrays.
Difference between two angles
In particular, the problem is when it comes to comparing angles. If you just take the difference of two close angles that are on the border between -pi and pi, you can get a value that is around 2*pi. The allclose just takes differences between values and checks that they are bellow some threshold. Thus in our cases, it can report a false negative.
A better way to compare angles is something along the lines of this function:
def angle_difference(a, b):
diff = a - b
diff[diff < -np.pi] += 2*np.pi
diff[diff > np.pi] -= 2*np.pi
return diff
You can then take the maximum absolute value and check that it's bellow some threshold:
np.max(np.abs(angle_difference(np.angle(mA), np.angle(rAfft)))) < threshold
In the case of your example, the maximum difference was 3.072209153742733e-12.
So the angles are actually correct!
Magnitude scaling
We can get an idea of the issue is when we look at the magnitude ratio between the matrix iDFT and the library iFFT.
print(np.abs(mA)/np.abs(rAfft))
We find that all the values in mA are 800, which means that our absolute values are 800 times larger than those computed by the library. Suspiciously, 800 = 40 * 20, the dimensions of our data! I think you can see where I am going with this.
Confusing DFT normalization
We spot some indications why this is the case when we have a look at the FFT formulas as taken from the Numpy FFT documentation:
You will notice that while the forward transform doesn't normalize by anything. The reverse transform divides the output by 1/N. These are the 1D FFTs but the exact same thing applies in the 2D case, the inverse transform multiplies everything by 1/(N*M)
So in our example, if we update this line, we will get the magnitudes to agree:
mA = dftMtxM # rA/(sizeM * sizeN) # dftMtxN
A side note on comparing the outputs, an alternative way to compare complex numbers is to compare the real and imaginary components:
print(np.allclose(mA.real, rAfft.real))
print(np.allclose(mA.imag, rAfft.imag))
And we find that now indeed both methods agree.
Why all this normalization mess and which should I use?
The fundamental property of the DFT transform must satisfy is that iDFT(DFT(x)) = x. When you work through the math, you find that the product of the two coefficients before the sum has to be 1/N.
There is also something called the Parseval's theorem. In simple terms, it states that the energy in the signals is just the sum of square absolutes in both the time domain and frequency domain. For the FFT this boils down to this relationship:
Here is the function for computing the energy of a signal:
def energy(x):
return np.sum(np.abs(x)**2)
You are basically faced with a choice about the 1/N factor:
You can put the 1/N before the DFT sum. This makes senses as then the k=0 DC component will be equal to the average of the time domain values. However you will have to multiply the energy in frequency domain by N in order to match it with time domain frequency.
N = len(x)
X = np.fft.fft(x)/N # Compute the FFT scaled by `1/N`
# Energy related by `N`
np.allclose(energy(x), energy(X) * N) == True
# Perform some processing...
Y = X * H
y = np.fft.ifft(Y*N) # Compute the iFFT, remember to cancel out the built in `1/N` of ifft
You put the 1/N before the iDFT. This is, slightly counterintuitively, what most implementations, including Numpy do. I could not find a definitive consensus on the reasoning behind this, but I think it has something to do with the implementation efficiency. (If anyone has a better explanation for this, please leave it in the comments) As shown in the equations earlier, the energy in the frequency domain has to be divided by N to match the time domain energy.
N = len(x)
X = np.fft.fft(x) # Compute the FFT without scaling
# Energy, related by 1/N
np.allclose(energy(x), energy(X) / N) == True
# Perform some processing...
Y = X * H
y = np.fft.ifft(Y) # Compute the iFFT with the build in `1/N`
You can split the 1/N by placing 1/sqrt(N) before each of the transforms making them perfectly symmetric. In Numpy, you can provide the parameter norm="ortho" to the fft functions which will make them use the 1/sqrt(N) normalization instead: np.fft.fft(x, norm="ortho") The nice property here is that the energy now matches in both domains.
X = np.fft.fft(x, norm='orth') # Compute the FFT scaled by `1/sqrt(N)`
# Perform some processing...
# Energy are equal:
np.allclose(energy(x), energy(X)) == True
Y = X * H
y = np.fft.ifft(Y, norm='orth') # Compute the iFFT, with scaling by `1/sqrt(N)`
In the end it boils down to what you need. Most of the time the absolute magnitude of your DFT is actually not that important. You are mostly interested in the ratio of various components or you want to perform some operation in the frequency domain but then transform back to the time domain or you are interested in the phase (angles). In all of these case, the normalization does not really play an important role, as long as you stay consistent.

equation system with fsolve

I try to find a solution for a system of equations by using scipy.optimize.fsolve in python 2.7. The goal is to calculate equilibrium concentrations for a chemical system. Due to the nature of the problem, some of the constants are very small. Now for some combinations i do get a proper solution. For some parameters i don't find a solution. Either the solutions are negative, which is not reasonable from a physical point of view or fsolve produces:
ier = 3, 'xtol=0.000000 is too small, no further improvement in the approximate\n solution is possible.')
ier = 4, 'The iteration is not making good progress, as measured by the \n improvement from the last five Jacobian evaluations.')
ier = 5, 'The iteration is not making good progress, as measured by the \n improvement from the last ten iterations.')
It seems to me, based on my research, that the failure to find proper solutions of the equation system is connected to the datatype float.64 not being precise enough. As a friend pointed out, the system is not well conditioned with parameters differing in several magnitudes.
So i tried to use fsolve with the mpfr type provided by the gmpy2 module but that resulted in the following error:
TypeError: Cannot cast array data from dtype('O') to dtype('float64') according to the rule 'safe'
Now here is a small example with parameter which lead to a solution if the randomized starting parameters fit happen to be good. However if the constant C_HCL is chosen to be something like 1e-4 or bigger then i never find a proper solution.
from numpy import *
from scipy.optimize import *
K_1 = 1e-8
K_2 = 1e-8
K_W = 1e-30
C_HCL = 1e-11
C_NAOH = K_W/C_HCL
C_HL = 1e-6
if C_HCL-C_NAOH > 0:
Saeure_Base = C_HCL-C_NAOH+sqrt(K_W)
OH_init = K_W/(Saeure_Base)
elif C_HCL-C_NAOH < 0:
OH_init = C_NAOH-C_HCL+sqrt(K_W)
Saeure_Base = K_W/OH_init
# some randomized start parameters
G1 = random.uniform(0, 2)*Saeure_Base
G2 = random.uniform(0, 2)*OH_init
G3 = random.uniform(1, 2)*C_HL*(sqrt(K_W))/(Saeure_Base+OH_init)
G4 = random.uniform(0.1, 1)*(C_HL - G3)/2
G5 = C_HL - G3 - G4
zGuess = array([G1,G2,G3,G4,G5])
#equation system / 5 variables --> H3O, OH, HL, H2L, L
def myFunction(z):
H3O = z[0]
OH = z[1]
HL = z[2]
H2L = z[3]
L = z[4]
F = empty((5))
F[0] = H3O*L/HL - K_1
F[1] = OH*H2L/HL - K_2
F[2] = K_W - OH*H3O
F[3] = C_HL - HL - H2L - L
F[4] = OH+L+C_HCL-H2L-H3O-C_NAOH
return F
z = fsolve(myFunction,zGuess, maxfev=10000, xtol=1e-15, full_output=1,factor=0.1)
print z
So the questions are. Is this problem based on the precision of float.64 and
if yes , (how) can it be solved with python? Is fsolve the way to go? Would i need to change the fsolve function so it accepts a different data type?

The root of your problem is either theoretical or numerical.
The scipy.optimize.fsolvefunction is based on the MINPACK Fortran solver (http://www.netlib.org/minpack/). This solver use a Newton-Raphson optimisation algorithm to provide the solution.
There are underlying assumptions about the smoothness of the function when you use this algorithm. For example, the jacobian matrix at the solution point x is supposed to be invertible. The one you are more concerned about is the basins of attraction.
In order to converge, the starting point of the algorithm needs to be near the actual solution, i.e. in the basins of attraction. This condition is always met for convex functions, however it is easy to find some functions for which this algorithm behaves badly. Your function is one of this as you have a fraction of your inputs parameters.
To address this issue you should just change the starting point. This starting point becomes also very important for functions with multiple solutions: this picture from the wikipedia article shows you the solution found depending of the starting point (five colours for five solutions); so you should be careful with your solution and actually check the "physical" aspects of your solution.
For the numerical aspects, the Newton-Raphson algorithm needs to have the value of the jacobian matrix (the derivatives matrix). If it is not provided to the MINPACK solver, the jacobian is estimated with a finite-difference formula. The perturbation step for the finite difference formula need to be provided epsfcn=None, the None being here as default value only in the case where fprimeis provided (there is no need for the jacobian estimation in this case). So first you should incorporate that. You could also specify directly the jacobian by derivating your function by hand.
However, the minimum value for the step size will be the machine precision, also called machine epsilon. For your problem, you have very small inputs values which can be a problem. I would suggest multiply everyone of them by the same value (like 10^6), it is equivalent to a change of the units but will avoid rounding up errors and problems with machine precision.
This problem is also important when you look at the parameter xtol=1e-15 you provided. In your error message, it gives xtol=0.000000, as it is below machine precision and cannot be taken into account. Also, if you look at your line F[2] = K_W - OH*H3O, given the machine precision, it does not matter if K_W is 1e-15or 1e-30. 0 is a solution for both of this case compare to the machine precision. To avoid this problem, just multiply everything by a bigger value.
So to sum up:
For the Newton-Raphson algorithm, the initialisation point matters !
For this algorithm, you should specify how you compute the jacobian !
In numerical computation, never work with small values. You can easily change the dimension to something different: it is basic units conversion, like working in gram instead of kilogram.

Fitting curve: why small numbers are better?

I spent some time these days on a problem. I have a set of data:
y = f(t), where y is very small concentration (10^-7), and t is in second. t varies from 0 to around 12000.
The measurements follow an established model:
y = Vs * t - ((Vs - Vi) * (1 - np.exp(-k * t)) / k)
And I need to find Vs, Vi, and k. So I used curve_fit, which returns the best fitting parameters, and I plotted the curve.
And then I used a similar model:
y = (Vs * t/3600 - ((Vs - Vi) * (1 - np.exp(-k * t/3600)) / k)) * 10**7
By doing that, t is a number of hour, and y is a number between 0 and about 10. The parameters returned are of course different. But when I plot each curve, here is what I get:
http://i.imgur.com/XLa4LtL.png
The green fit is the first model, the blue one with the "normalized" model. And the red dots are the experimental values.
The fitting curves are different. I think it's not expected, and I don't understand why. Are the calculations more accurate if the numbers are "reasonnable" ?

The docstring for optimize.curve_fit says,
p0 : None, scalar, or M-length sequence
Initial guess for the parameters. If None, then the initial
values will all be 1 (if the number of parameters for the function
can be determined using introspection, otherwise a ValueError
is raised).
Thus, to begin with, the initial guess for the parameters is by default 1.
Moreover, curve fitting algorithms have to sample the function for various values of the parameters. The "various values" are initially chosen with an initial step size on the order of 1. The algorithm will work better if your data varies somewhat smoothly with changes in the parameter values that on the order of 1.
If the function varies wildly with parameter changes on the order of 1, then the algorithm may tend to miss the optimum parameter values.
Note that even if the algorithm uses an adaptive step size when it tweaks the parameter values, if the initial tweak is so far off the mark as to produce a big residual, and if tweaking in some other direction happens to produce a smaller residual, then the algorithm may wander off in the wrong direction and miss the local minimum. It may find some other (undesired) local minimum, or simply fail to converge. So using an algorithm with an adaptive step size won't necessarily save you.
The moral of the story is that scaling your data can improve the algorithm's chances of of finding the desired minimum.
Numerical algorithms in general all tend to work better when applied to data whose magnitude is on the order of 1. This bias enters into the algorithm in numerous ways. For instance, optimize.curve_fit relies on optimize.leastsq, and the call signature for optimize.leastsq is:
def leastsq(func, x0, args=(), Dfun=None, full_output=0,
col_deriv=0, ftol=1.49012e-8, xtol=1.49012e-8,
gtol=0.0, maxfev=0, epsfcn=None, factor=100, diag=None):
Thus, by default, the tolerances ftol and xtol are on the order of 1e-8. If finding the optimum parameter values require much smaller tolerances, then these hard-coded default numbers will cause optimize.curve_fit to miss the optimize parameter values.
To make this more concrete, suppose you were trying to minimize f(x) = 1e-100*x**2. The factor of 1e-100 squashes the y-values so much that a wide range of x-values (the parameter values mentioned above) will fit within the tolerance of 1e-8. So, with un-ideal scaling, leastsq will not do a good job of finding the minimum.
Another reason to use floats on the order of 1 is because there are many more (IEEE754) floats in the interval [-1,1] than there are far away from 1. For example,
import struct
def floats_between(x, y):
"""
http://stackoverflow.com/a/3587987/190597 (jsbueno)
"""
a = struct.pack("<dd", x, y)
b = struct.unpack("<qq", a)
return b[1] - b[0]
In [26]: floats_between(0,1) / float(floats_between(1e6,1e7))
Out[26]: 311.4397707054894
This shows there are over 300 times as many floats representing numbers between 0 and 1 than there are in the interval [1e6, 1e7].
Thus, all else being equal, you'll typically get a more accurate answer if working with small numbers than very large numbers.

I would imagine it has more to do with the initial parameter estimates you are passing to curve fit. If you are not passing any I believe they all default to 1. Normalizing your data makes those initial estimates closer to the truth. If you don't want to use normalized data just pass the initial estimates yourself and give them reasonable values.

Others have already mentioned that you probably need to have a good starting guess for your fit. In cases like this is, I usually try to find some quick and dirty tricks to get at least a ballpark estimate of the parameters. In your case, for large t, the exponential decays pretty quickly to zero, so for large t, you have
y == Vs * t - (Vs - Vi) / k
Doing a first-order linear fit like
[slope1, offset1] = polyfit(t[t > 2000], y[t > 2000], 1)
you will get slope1 == Vs and offset1 == (Vi - Vs) / k.
Subtracting this straight line from all the points you have, you get the exponential
residual == y - slope1 * t - offset1 == (Vs - Vi) * exp(-t * k)
Taking the log of both sides, you get
log(residual) == log(Vs - Vi) - t * k
So doing a second fit
[slope2, offset2] = polyfit(t, log(y - slope1 * t - offset1), 1)
will give you slope2 == -k and offset2 == log(Vs - Vi), which should be solvable for Vi since you already know Vs. You might have to limit the second fit to small values of t, otherwise you might be taking the log of negative numbers. Collect all the parameters you obtained with these fits and use them as the starting points for your curve_fit.
Finally, you might want to look into doing some sort of weighted fit. The information about the exponential part of your curve is contained in just the first few points, so maybe you should give those a higher weight. Doing this in a statistically correct way is not trivial.