SciPy is large library used everywhere in scientific computing. That's why breaking backwards-compatibility comes as a significant cost and is almost always avoided, even if the API of some methods is arguably lacking. This package provides drop-in wrappers "fixing" those.

npx does the same for NumPy.

If you have a fix for a SciPy method that can't go upstream for some reason, feel free to PR here.

Krylov methods

import numpy as np
import scipy.sparse
import scipyx as spx

# create tridiagonal (-1, 2, -1) matrix
n = 100
data = -np.ones((3, n))
data[1] = 2.0
A = scipy.sparse.spdiags(data, [-1, 0, 1], n, n)
A = A.tocsr()
b = np.ones(n)

sol, info =, b, tol=1.0e-10)
sol, info = spx.minres(A, b, tol=1.0e-10)
sol, info = spx.gmres(A, b, tol=1.0e-10)
sol, info = spx.bicg(A, b, tol=1.0e-10)
sol, info = spx.bicgstab(A, b, tol=1.0e-10)
sol, info = spx.cgs(A, b, tol=1.0e-10)
sol, info = spx.qmr(A, b, tol=1.0e-10)

sol is the solution of the linear system A @ x = b (or None if no convergence),
and info contains some useful data, e.g., info.resnorms. The solution sol and all
callback x have the shape of x0/b.
The methods are wrappers around SciPy's iterative

Relevant issues:


import scipyx as spx

def f(x):
    return (x ** 2 - 2) ** 2

x0 = 1.5
out = spx.minimize(f, x0)

x0 = -3.2
x, _ = spx.leastsq(f, x0)

In scipyx, all intermediate values x and the result from a minimization out.x will
have the same shape as x0. (In SciPy, they always have shape (n,), no matter the
input vector.)

Relevant issues:


import scipyx as spx

def f(x):
    return x ** 2 - 2

a, b = spx.bisect(f, 0.0, 5.0, tol=1.0e-12)
a, b = spx.regula_falsi(f, 0.0, 5.0, tol=1.0e-12)

scipyx provides some basic nonlinear root-findings algorithms:
bisection and regula
. They're not as fast-converging as
other methods, but are very robust
and work with almost any function.