110 lines
2.7 KiB
Python
110 lines
2.7 KiB
Python
|
"""Numerical Methods for Holonomic Functions"""
|
||
|
|
|
||
|
|
from sympy.core.sympify import sympify
|
||
|
|
from sympy.holonomic.holonomic import DMFsubs
|
||
|
|
|
||
|
|
from mpmath import mp
|
||
|
|
|
||
|
|
|
||
|
|
def _evalf(func, points, derivatives=False, method='RK4'):
|
||
|
|
"""
|
||
|
|
Numerical methods for numerical integration along a given set of
|
||
|
|
points in the complex plane.
|
||
|
|
"""
|
||
|
|
|
||
|
|
ann = func.annihilator
|
||
|
|
a = ann.order
|
||
|
|
R = ann.parent.base
|
||
|
|
K = R.get_field()
|
||
|
|
|
||
|
|
if method == 'Euler':
|
||
|
|
meth = _euler
|
||
|
|
else:
|
||
|
|
meth = _rk4
|
||
|
|
|
||
|
|
dmf = []
|
||
|
|
for j in ann.listofpoly:
|
||
|
|
dmf.append(K.new(j.rep))
|
||
|
|
|
||
|
|
red = [-dmf[i] / dmf[a] for i in range(a)]
|
||
|
|
|
||
|
|
y0 = func.y0
|
||
|
|
if len(y0) < a:
|
||
|
|
raise TypeError("Not Enough Initial Conditions")
|
||
|
|
x0 = func.x0
|
||
|
|
sol = [meth(red, x0, points[0], y0, a)]
|
||
|
|
|
||
|
|
for i, j in enumerate(points[1:]):
|
||
|
|
sol.append(meth(red, points[i], j, sol[-1], a))
|
||
|
|
|
||
|
|
if not derivatives:
|
||
|
|
return [sympify(i[0]) for i in sol]
|
||
|
|
else:
|
||
|
|
return sympify(sol)
|
||
|
|
|
||
|
|
|
||
|
|
def _euler(red, x0, x1, y0, a):
|
||
|
|
"""
|
||
|
|
Euler's method for numerical integration.
|
||
|
|
From x0 to x1 with initial values given at x0 as vector y0.
|
||
|
|
"""
|
||
|
|
|
||
|
|
A = sympify(x0)._to_mpmath(mp.prec)
|
||
|
|
B = sympify(x1)._to_mpmath(mp.prec)
|
||
|
|
y_0 = [sympify(i)._to_mpmath(mp.prec) for i in y0]
|
||
|
|
h = B - A
|
||
|
|
f_0 = y_0[1:]
|
||
|
|
f_0_n = 0
|
||
|
|
|
||
|
|
for i in range(a):
|
||
|
|
f_0_n += sympify(DMFsubs(red[i], A, mpm=True))._to_mpmath(mp.prec) * y_0[i]
|
||
|
|
f_0.append(f_0_n)
|
||
|
|
|
||
|
|
sol = []
|
||
|
|
for i in range(a):
|
||
|
|
sol.append(y_0[i] + h * f_0[i])
|
||
|
|
|
||
|
|
return sol
|
||
|
|
|
||
|
|
|
||
|
|
def _rk4(red, x0, x1, y0, a):
|
||
|
|
"""
|
||
|
|
Runge-Kutta 4th order numerical method.
|
||
|
|
"""
|
||
|
|
|
||
|
|
A = sympify(x0)._to_mpmath(mp.prec)
|
||
|
|
B = sympify(x1)._to_mpmath(mp.prec)
|
||
|
|
y_0 = [sympify(i)._to_mpmath(mp.prec) for i in y0]
|
||
|
|
h = B - A
|
||
|
|
|
||
|
|
f_0_n = 0
|
||
|
|
f_1_n = 0
|
||
|
|
f_2_n = 0
|
||
|
|
f_3_n = 0
|
||
|
|
|
||
|
|
f_0 = y_0[1:]
|
||
|
|
for i in range(a):
|
||
|
|
f_0_n += sympify(DMFsubs(red[i], A, mpm=True))._to_mpmath(mp.prec) * y_0[i]
|
||
|
|
f_0.append(f_0_n)
|
||
|
|
|
||
|
|
f_1 = [y_0[i] + f_0[i]*h/2 for i in range(1, a)]
|
||
|
|
for i in range(a):
|
||
|
|
f_1_n += sympify(DMFsubs(red[i], A + h/2, mpm=True))._to_mpmath(mp.prec) * (y_0[i] + f_0[i]*h/2)
|
||
|
|
f_1.append(f_1_n)
|
||
|
|
|
||
|
|
f_2 = [y_0[i] + f_1[i]*h/2 for i in range(1, a)]
|
||
|
|
for i in range(a):
|
||
|
|
f_2_n += sympify(DMFsubs(red[i], A + h/2, mpm=True))._to_mpmath(mp.prec) * (y_0[i] + f_1[i]*h/2)
|
||
|
|
f_2.append(f_2_n)
|
||
|
|
|
||
|
|
f_3 = [y_0[i] + f_2[i]*h for i in range(1, a)]
|
||
|
|
for i in range(a):
|
||
|
|
f_3_n += sympify(DMFsubs(red[i], A + h, mpm=True))._to_mpmath(mp.prec) * (y_0[i] + f_2[i]*h)
|
||
|
|
f_3.append(f_3_n)
|
||
|
|
|
||
|
|
sol = []
|
||
|
|
for i in range(a):
|
||
|
|
sol.append(y_0[i] + h * (f_0[i]+2*f_1[i]+2*f_2[i]+f_3[i])/6)
|
||
|
|
|
||
|
|
return sol
|