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.

Overflow in numpy.exp()

I have to calculate the exponential of the following array for my project:
w = [-1.52820754859, -0.000234000845064, -0.00527938881237, 5797.19232191, -6.64682108484,
18924.7087966, -69.308158911, 1.1158892974, 1.04454511882, 116.795573742]
But I've been getting overflow due to the number 18924.7087966.
The goal is to avoid using extra packages such as bigfloat (except "numpy") and get a close result (which has a small relative error).
1.So far I've tried using higher precision (i.e. float128):
def getlogZ_robust(w):
Z = sum(np.exp(np.dot(x,w).astype(np.float128)) for x in iter_all_observations())
return np.log(Z)
But I still get "inf" which is what I want to avoid.
I've tried clipping it using nump.clip():
def getlogZ_robust(w):
Z = sum(np.exp(np.clip(np.dot(x,w).astype(np.float128),-11000, 11000)) for x in iter_all_observations())
return np.log(Z)
But the relative error is too big.
Can you help me solving this problem, if it is possible?

Only significantly extended or arbitrary precision packages will be able to handle the huge differences in numbers. The exponential of the largest and most negative numbers in w differ by 8000 (!) orders of magnitude. float (i.e. double precision) has 'only' 15 digits of precision (meaning 1+1e-16 is numerically equal to 1), such that adding the small numbers to the huge exponential of the largest number has no effect. As a matter of fact, exp(18924.7087966) is so huge, that it dominates the sum. Below is a script performing the sum with extended precision in mpmath: the ratio of the sum of exponentials and exp(18924.7087966) is basically 1.
w = [-1.52820754859, -0.000234000845064, -0.00527938881237, 5797.19232191, -6.64682108484,
18924.7087966, -69.308158911, 1.1158892974, 1.04454511882, 116.795573742]
u = min(w)
v = max(w)
import mpmath
#using plenty of precision
mpmath.mp.dps = 32768
print('%.5e' % mpmath.log10(mpmath.exp(v)/mpmath.exp(u)))
#exp(w) differs by 8000 orders of magnitude for largest and smallest number
s = sum([mpmath.exp(mpmath.mpf(x)) for x in w])
print('%.5e' % (mpmath.exp(v)/s))
#largest exp(w) dominates such that ratio over the sums of exp(w) and exp(max(w)) is approx. 1

If the issues of loosing digits in the final results due to hugely different orders of magnitudes of added terms in not a concern, one could also mathematically transform the log of sums over exponentials the following way avoiding exp of large numbers:
log(sum(exp(w)))
= log(sum(exp(w-wmax)*exp(wmax)))
= wmax + log(sum(exp(w-wmax)))
In python:
import numpy as np
v = np.array(w)
m = np.max(v)
print(m + np.log(np.sum(np.exp(v-m))))
Note that np.log(np.sum(np.exp(v-m))) is numerically zero as the exponential of the largest number completely dominates the sum here.

Numpy has a function called logaddexp which computes
logaddexp(x1, x2) == log(exp(x1) + exp(x2))
without explicitly computing the intermediate exp() values. This way it avoids the overflow. So here is the solution:
def getlogZ_robust(w):
Z = 0
for x in iter_all_observations():
Z = np.logaddexp(Z, np.dot(x,w))
return Z

How do I find the percentage error in a Monte Carlo algorithm?

I have written a Monte Carlo program to integrate a function f(x).
I have now been asked to calculate the percentage error.
Having done a quick literature search, I found that this can be given with the equation %error = (sqrt(var[f(x)]/n))*100, where n is the number of random points I used to derive my answer.
However, when I run my integration code, my percentage error is greater than that given by this formula.
Do I have the correct formula?
Any help would be greatly appreciated. Thanks x

Here is quick example - estimate integral of linear function on the interval [0...1] using Monte-Carlo. To estimate error you have to collect second momentum (values squared), then compute variance, standard deviation, and (assuming CLT), error of the simulation in the original units as well as in %
Code, Python 3.7, Anaconda, Win10 64x
import numpy as np
def f(x): # linear function to integrate
return x
np.random.seed(312345)
N = 100000
x = np.random.random(N)
q = f(x) # first momentum
q2 = q*q # second momentum
mean = np.sum(q) / float(N) # compute mean explicitly, not using np.mean
var = np.sum(q2) / float(N) - mean * mean # variance as E[X^2] - E[X]^2
sd = np.sqrt(var) # std.deviation
print(mean) # should be 1/2
print(var) # should be 1/12
print(sd) # should be 0.5/sqrt(3)
print("-----------------------------------------------------")
sigma = sd / np.sqrt(float(N)) # assuming CLT, error estimation in original units
print("result = {0} with error +- {1}".format(mean, sigma))
err_pct = sigma / mean * 100.0 # error estimate in percents
print("result = {0} with error +- {1}%".format(mean, err_pct))
Be aware, that we computed one sigma error and (even not talking about it being random value itself) true result is within printed mean+-error only for 68% of the runs. You could print mean+-2*error, and it would mean true result is inside that region for 95% cases, mean+-3*error true result is inside that region for 99.7% of the runs and so on and so forth.
UPDATE
For sampling variance estimate, there is known problem called Bias in the estimator. Basically, we underestimate a bit sampling variance, proper correction (Bessel's correction) shall be applied
var = np.sum(q2) / float(N) - mean * mean # variance as E[X^2] - E[X]^2
var *= float(N)/float(N-1)
In many cases (and many examples) it is omitted because N is very large, which makes correction pretty much invisible - f.e., if you have statistical error 1% but N is in millions, correction is of no practical use.

Why is my code using 4th Runge-Kutta isn't giving me the expected values?

I'm having a little trouble trying to understand what's wrong with me code, any help would be extremely helpful.
I wanted to solve this simple equation
However, the values my code gives doesn't match with my book ones or wolfram ones as y goes up as x grows.
import matplotlib.pyplot as plt
from numpy import exp
from scipy.integrate import ode
# initial values
y0, t0 = [1.0], 0.0
def f(t, y):
f = [3.0*y[0] - 4.0/exp(t)]
return f
# initialize the 4th order Runge-Kutta solver
r = ode(f).set_integrator('dopri5')
r.set_initial_value(y0, t0)
t1 = 10
dt = 0.1
x, y = [], []
while r.successful() and r.t < t1:
x.append(r.t+dt); y.append(r.integrate(r.t+dt))
print(r.t+dt, r.integrate(r.t+dt))

Your equation in general has the solution
y(x) = (y0-1)*exp(3*x) + exp(-x)
Due to the choice of initial conditions, the exact solution does not contain the growing component of the first term. However, small perturbations due to discretization and floating point errors will generate a non-zero coefficient in the growing term. Now at the end of the integration interval this random coefficient is multiplied by exp(3*10)=1.107e+13 which will magnify small discretization errors of size 1e-7 to contributions in the result of size 1e+6 as observed when running the original code.
You can force the integrator to be more precise in its internal steps without reducing the output step size dt by setting error thresholds like in
r = ode(f).set_integrator('dopri5', atol=1e-16, rtol=1e-20)
However, you can not avoid the deterioration of the result completely as the floating point errors of size 1e-16 get magnified to global error contributions of size 1e-3.
Also, you should notice that each call of r.integrate(r.t+dt) will advance the integrator by dt so that the stored array and the printed values are in lock-step. If you want to just print the current state of the integrator use
print(r.t,r.y,yexact(r.t,y0))
where the last is to compare to the exact solution which is, as already said,
def yexact(x,y0):
return [ (y0[0]-1)*exp(3*x)+exp(-x) ]

How to find all zeros of a function using numpy (and scipy)?

Suppose I have a function f(x) defined between a and b. This function can have many zeros, but also many asymptotes. I need to retrieve all the zeros of this function. What is the best way to do it?
Actually, my strategy is the following:
I evaluate my function on a given number of points
I detect whether there is a change of sign
I find the zero between the points that are changing sign
I verify if the zero found is really a zero, or if this is an asymptote
U = numpy.linspace(a, b, 100) # evaluate function at 100 different points
c = f(U)
s = numpy.sign(c)
for i in range(100-1):
if s[i] + s[i+1] == 0: # oposite signs
u = scipy.optimize.brentq(f, U[i], U[i+1])
z = f(u)
if numpy.isnan(z) or abs(z) > 1e-3:
continue
print('found zero at {}'.format(u))
This algorithm seems to work, except I see two potential problems:
It will not detect a zero that doesn't cross the x axis (for example, in a function like f(x) = x**2) However, I don't think it can occur with the function I'm evaluating.
If the discretization points are too far, there could be more that one zero between them, and the algorithm could fail finding them.
Do you have a better strategy (still efficient) to find all the zeros of a function?
I don't think it's important for the question, but for those who are curious, I'm dealing with characteristic equations of wave propagation in optical fiber. The function looks like (where V and ell are previously defined, and ell is an positive integer):
def f(u):
w = numpy.sqrt(V**2 - u**2)
jl = scipy.special.jn(ell, u)
jl1 = scipy.special.jnjn(ell-1, u)
kl = scipy.special.jnkn(ell, w)
kl1 = scipy.special.jnkn(ell-1, w)
return jl / (u*jl1) + kl / (w*kl1)

Why are you limited to numpy? Scipy has a package that does exactly what you want:
http://docs.scipy.org/doc/scipy/reference/optimize.nonlin.html
One lesson I've learned: numerical programming is hard, so don't do it :)
Anyway, if you're dead set on building the algorithm yourself, the doc page on scipy I linked (takes forever to load, btw) gives you a list of algorithms to start with. One method that I've used before is to discretize the function to the degree that is necessary for your problem. (That is, tune \delta x so that it is much smaller than the characteristic size in your problem.) This lets you look for features of the function (like changes in sign). AND, you can compute the derivative of a line segment (probably since kindergarten) pretty easily, so your discretized function has a well-defined first derivative. Because you've tuned the dx to be smaller than the characteristic size, you're guaranteed not to miss any features of the function that are important for your problem.
If you want to know what "characteristic size" means, look for some parameter of your function with units of length or 1/length. That is, for some function f(x), assume x has units of length and f has no units. Then look for the things that multiply x. For example, if you want to discretize cos(\pi x), the parameter that multiplies x (if x has units of length) must have units of 1/length. So the characteristic size of cos(\pi x) is 1/\pi. If you make your discretization much smaller than this, you won't have any issues. To be sure, this trick won't always work, so you may need to do some tinkering.

I found out it's relatively easy to implement your own root finder using the scipy.optimize.fsolve.
Idea: Find any zeroes from interval (start, stop) and stepsize step by calling the fsolve repeatedly with changing x0. Use relatively small stepsize to find all the roots.
Can only search for zeroes in one dimension (other dimensions must be fixed). If you have other needs, I would recommend using sympy for calculating the analytical solution.
Note: It may not always find all the zeroes, but I saw it giving relatively good results. I put the code also to a gist, which I will update if needed.
import numpy as np
import scipy
from scipy.optimize import fsolve
from matplotlib import pyplot as plt
# Defined below
r = RootFinder(1, 20, 0.01)
args = (90, 5)
roots = r.find(f, *args)
print("Roots: ", roots)
# plot results
u = np.linspace(1, 20, num=600)
fig, ax = plt.subplots()
ax.plot(u, f(u, *args))
ax.scatter(roots, f(np.array(roots), *args), color="r", s=10)
ax.grid(color="grey", ls="--", lw=0.5)
plt.show()
Example output:
Roots: [ 2.84599497 8.82720551 12.38857782 15.74736542 19.02545276]
zoom-in:
RootFinder definition
import numpy as np
import scipy
from scipy.optimize import fsolve
from matplotlib import pyplot as plt
class RootFinder:
def __init__(self, start, stop, step=0.01, root_dtype="float64", xtol=1e-9):
self.start = start
self.stop = stop
self.step = step
self.xtol = xtol
self.roots = np.array([], dtype=root_dtype)
def add_to_roots(self, x):
if (x < self.start) or (x > self.stop):
return # outside range
if any(abs(self.roots - x) < self.xtol):
return # root already found.
self.roots = np.append(self.roots, x)
def find(self, f, *args):
current = self.start
for x0 in np.arange(self.start, self.stop + self.step, self.step):
if x0 < current:
continue
x = self.find_root(f, x0, *args)
if x is None: # no root found.
continue
current = x
self.add_to_roots(x)
return self.roots
def find_root(self, f, x0, *args):
x, _, ier, _ = fsolve(f, x0=x0, args=args, full_output=True, xtol=self.xtol)
if ier == 1:
return x[0]
return None
Test function
The scipy.special.jnjn does not exist anymore, but I created similar test function for the case.
def f(u, V=90, ell=5):
w = np.sqrt(V ** 2 - u ** 2)
jl = scipy.special.jn(ell, u)
jl1 = scipy.special.yn(ell - 1, u)
kl = scipy.special.kn(ell, w)
kl1 = scipy.special.kn(ell - 1, w)
return jl / (u * jl1) + kl / (w * kl1)

The main problem I see with this is if you can actually find all roots --- as have already been mentioned in comments, this is not always possible. If you are sure that your function is not completely pathological (sin(1/x) was already mentioned), the next one is what's your tolerance to missing a root or several of them. Put differently, it's about to what length you are prepared to go to make sure you did not miss any --- to the best of my knowledge, there is no general method to isolate all the roots for you, so you'll have to do it yourself. What you show is a reasonable first step already. A couple of comments:
Brent's method is indeed a good choice here.
First of all, deal with the divergencies. Since in your function you have Bessels in the denominators, you can first solve for their roots -- better look them up in e.g., Abramovitch and Stegun (Mathworld link). This will be a better than using an ad hoc grid you're using.
What you can do, once you've found two roots or divergencies, x_1 and x_2, run the search again in the interval [x_1+epsilon, x_2-epsilon]. Continue until no more roots are found (Brent's method is guaranteed to converge to a root, provided there is one).
If you cannot enumerate all the divergencies, you might want to be a little more careful in verifying a candidate is indeed a divergency: given x don't just check that f(x) is large, check that, e.g. |f(x-epsilon/2)| > |f(x-epsilon)| for several values of epsilon (1e-8, 1e-9, 1e-10, something like that).
If you want to make sure you don't have roots which simply touch zero, look for the extrema of the function, and for each extremum, x_e, check the value of f(x_e).

I've also encountered this problem to solve equations like f(z)=0 where f was an holomorphic function. I wanted to be sure not to miss any zero and finally developed an algorithm which is based on the argument principle.
It helps to find the exact number of zeros lying in a complex domain. Once you know the number of zeros, it is easier to find them. There are however two concerns which must be taken into account :
Take care about multiplicity : when solving (z-1)^2 = 0, you'll get two zeros as z=1 is counting twice
If the function is meromorphic (thus contains poles), each pole reduce the number of zero and break the attempt to count them.