Jeff Epler
196104084b
Mostly by providing a "numpy shim" for CircuitPython, try to make the numpy tests run on all three systems. (a "scipy shim" might also be useful?) However, there are test failures. Is it worth working through them and getting this to a point where it could be included?
37 lines
1.1 KiB
Python
37 lines
1.1 KiB
Python
import math
|
|
try:
|
|
import scipy.optimize as optimize
|
|
except:
|
|
try:
|
|
import ulab.approx as optimize
|
|
except:
|
|
import ulab as optimize
|
|
|
|
try:
|
|
from math import isclose
|
|
except:
|
|
def isclose(a, b, *, rel_tol=1e-9, abs_tol=0):
|
|
print("isclose", a, b);
|
|
return abs(a-b) <= max(rel_tol * max(abs(a), abs(b)), abs_tol)
|
|
|
|
def f(x):
|
|
return x**2 - 2.0
|
|
|
|
ref_result = 1.41421826342255
|
|
result = optimize.newton(f, 3., tol=0.001, rtol=0.01)
|
|
print(ref_result, result)
|
|
print(isclose(result, ref_result, rel_tol=1E-9, abs_tol=1E-9))
|
|
result = optimize.newton(f, 3., tol=0.001, rtol=0.01, maxiter=100)
|
|
print(isclose(result, ref_result, rel_tol=1E-9, abs_tol=1E-9))
|
|
|
|
ref_result = 1.4142135623715149
|
|
result = optimize.bisect(f, 1.0, 3.0)
|
|
print(isclose(result, ref_result, rel_tol=1E-9, abs_tol=1E-9))
|
|
|
|
#ref_result = -7.105427357601002e-15
|
|
#result = optimize.fmin(f, 3.0, disp=0)
|
|
#print(isclose(result[0], ref_result, rel_tol=1E-9, abs_tol=1E-9))
|
|
#result = optimize.fmin(f, 3.0, xtol=0.0001, ftol=0.0001, disp=0)
|
|
#print(isclose(result[0], ref_result, rel_tol=1E-9, abs_tol=1E-9))
|
|
|
|
|