def taylor(x,f,i,n):
"""taylor(x,f,i,n):
This function approximates the function f over the domain x,
using a taylor expansion centered at x[i]
with n+1 terms (starts counting from 0).
Args:
x: The domain of the function
f: The function that will be expanded/approximated
i: The ith term in the domain x that the expansion is centered around
n: The number of terms in the expansion (n+1 terms)
Returns:
(x,fapprox): A pair of numpy arrays where x is the original domain array and
f approx is the approximation of f over all of the domain points x using the
taylor expansion.
"""
a = x[i]
N = np.size(x)
fa = f[i]*np.ones_like(x)
D = ac.derivative(x[0],x[N-1],N)
fact = 1
fapprox = fa
Dk = np.eye(N)
for k in range(1,n+1):
fact = fact*k
Dk = np.matmul(Dk,D)
#fapprox += (np.matmul(np.matmul(Dk,fa),((x-a)**k)))/fact
fapprox = np.add(fapprox, (np.matmul(np.matmul(Dk,fa),((x-a)**k)))/fact, out=fapprox, casting="unsafe")
return (x,fapprox)
def taylor(x, f, i, n):
"""taylor(x,f,i,n)
Inputs:
-x: array of domain points
-f: array of range points for a function f(x)
-i: integer, between 0 and len(x), such that x[i] will be the point expanded around
-n: positive integer, number of terms kept in the approximation
Returns: (x, fapprox)
-x: same array of domain points as inputted
-fapprox: array of the taylor approximated function at point x[i]"""
length = len(x)
mat = ac.derivative(x[0], x[length - 1], length)
fapprox = np.zeros(length)
def mat_generator():
n_mat = np.eye(length)
while True:
n_mat = np.dot(n_mat, mat)
yield n_mat
g = mat_generator()
k = np.arange(length)
a = x[i]
fapprox[k] = f[i]
for j in range(1, n):
n_mat = next(g)
fapprox += (np.dot(n_mat, f)[i] * ((x - a)**j) /
(np.math.factorial(j)))
return (x, fapprox)
def test_derivative():
"""Tests derivative(a,b,n) with following tests:
- derivative(0,9,10) multiplied by f = [0,1,2,3,4,5,6,7,8,9]
"""
actual = np.array([1,1,1,1,1,1,1,1,1,1])
trial = ac.derivative(0,9,10)@(np.array([0,1,2,3,4,5,6,7,8,9]))
print("Testing: actual ?= trial: ",actual," += ",trial)
np.testing.assert_almost_equal(actual,trial)
def test_derivative():
"""
test_derivative function description:
Tests whether the central finite difference is close to the derivative of the function
"""
f = lambda x: x**2
Df = ac.derivative(f, xmax=99)
for i in range(1, len(Df) - 1):
assert math.isclose(Df[i], 2 * i)
def test_derivative():
"""test_derivative()
Tests for the first derivative accuracy
"""
a = 0
b = 3
n = 3
f = np.array([2.0, 2.0, 2.0])
answer = np.empty(n)
D1 = array_calculus.derivative(a, b, n)
answer = D1 @ f
desired = np.array([0.0, 0.0, 0.0])
np.testing.assert_almost_equal(answer, desired)
def test_derivative_2():
"""test_derivative_2()
Tests for the first derivative accuracy specifically for function x^2
"""
a = 0
b = 5
n = 5
#f(x)=x^2 and x_array=[(0,1,2,3,4)], which is why the f array is defined this way
f = np.array([0.0, 1.0, 4.0, 9.0, 16.0])
answer = np.empty(n)
D1 = array_calculus.derivative(a, b, n)
#D2 = array_calculus.second_derivative(a,b,n)
answer = D1 @ f
desired = np.array([0.8, 1.6, 3.2, 4.8, 5.6])
np.testing.assert_almost_equal(answer, desired)
def test_derivative():
"""Testing our derivative matrix for three point of an exponential function"""
# Testing for derivative of exponential at points x=0,1,2
actual = np.array([(np.exp(1)-np.exp(0)), (np.exp(2)-np.exp(0))/2, (np.exp(2)-np.exp(1))])
# Testing implementation
def exponential():
t = np.linspace(0,2,3)
ex = np.vectorize(np.exp)
ex = ex(t)
return ex
trial = np.dot(ac.derivative(0,2,3),exponential())
# Debug message
print("Should be: ",actual," but returned this: ",trial)
for m in range(3):
nose.tools.assert_almost_equal(actual[m],trial[m],4)
def test_derivative():
"""Test_derivative() tests the derivative matrix for the chosen values of
x= -1, 0, and 1."""
desired = np.array([(np.sin(0) - np.sin(-1)), (np.sin(1) - np.sin(-1) / 2),
(np.sin(1) - np.sin(0))])
def sin():
x = np.linspace(-1, 1, 3)
sin = np.sin(x)
return sin
obtained = np.array(ac.derivative(-1, 1, 3), sin())
print("Desired: ", desired)
print("Obtained: ", obtained)
np.testing.assert_almost_equal(obtained, desired)
def taylor(x, f, i, n):
y = len(x)
mat = ac.derivative(x[0], x[y - 1], y)
fx = np.zeros(y) #array on zeros of length y
def tay_matrix():
matrix = np.eye(y)
while True:
matrix = np.dot(matrix, mat)
yield matrix
yint = np.arange(y)
m = tay_matrix()
fx[yint] = f[i]
array = x[i]
for k in range(1, n):
matrix = next(m)
fx += (np.dot(matrix, f)[i] * ((x - array)**k) /
(np.math.factorial(k)))
return (x, fx)
def taylor(x, f, i, n):
'''taylor(x,f,i,n=10)
Takes the taylor series approximation around a specific point at position i.
Args:
x (array) : x-coordinates
f (array) : y-coordinates
i (index int) : position of point between 0 and the domain
n (int) : number of terms to keep in Taylor expansion sum
Returns:
(x, fapprox) : Pair of numpy arrays
x : x-coordinates domain
fapprox : new approximate function y-coordinates obtained by Taylor formula at x[i]
'''
fapprox = np.zeros(len(x))
for xValue in range(0, len(x)):
answer = 0
for term in range(0, n + 1):
D = array_calculus.derivative(x[0], x[len(x) - 1], len(x))
D_operator = np.linalg.matrix_power(D, term)
myDerivative = D_operator @ f
answer = answer + (myDerivative[i] * ((
(x[xValue] - x[i])**term) / math.factorial(term)))
fapprox[xValue] = answer
return (x, fapprox)
def test_first_and_Second_derivatives():
actual = calc.second_derivative(0,9,9)
trial = calc.derivative(0,9,9)@calc.derivative(0,9,9)
print("Testing first derivative: ",actual," ?= ",trial)
np.testing.assert_array_almost_equal(actual, trial)
