Generic derivative objects (gradients, Jacobians, Hessians, and more) by finite differences
Project description
taylor
Compute finite difference approximations to derivatives of multidimensional, multivariate functions with respect to multidimensional variables.
Unlike findiff
and fastfd
(which focus on the computation of finite differences over multidimensional grids likely representing physical systems), the focus of taylor
is on computing higher order derivative objects (gradients, Jacobians, Hessians, etc.) for problems like optimization and the solution of nonlinear equations.
Of course, it may also be used to numerically verify handcoded implementations of said derivative objects.
Interface
def diff(fun,x,order,args=(),mask=None,rule='forward',delta=None,
idx_order='default'):
"""
fun : {function}
function whose derivative is sought
has function definition
def fun(x,*args):
x : {scalar, array}
independent variable with respect to which the derivative
will be computed
order : {integer}
order of desired derivative
1: first derivative (gradient),
2: second derivative (Hessian), ...
args : {tuple}
tuple of additional arguments to fun
mask : {integer or array}
array of same shape as the returned derivative where
element = 1 -> this entry should be computed,
element = 0 -> entry should not be computed
rule : {string}
finite difference rule
choose from: {'forward','backward','central'}
delta : {float or array}
scalar/array of same shape as x that specifies the finite
difference step size
idx_order : {string}
string indicating how indices of derivative object should be
ordered when returned
'default' : indices corresponding to derivatives are
ordered first
'natural' : indices corresponding to elements of function
output are ordered first (like in Jacobians)
"""
Examples
The first example computes the first derivative of the matrix vector product f(A,x) = A x
with respect to both the matrix A
and vector x
.
import numpy as np
import taylor as ta
# both functions compute matrix vector product,
# but have different first arguments
def matvec_vec(x,A):
return A @ x
def matvec_mat(A,x):
return A @ x
if (__name__ == "__main__"):
# set matrix and vector
A = np.array([[1.0,2.0,3.0],
[2.0,4.0,5.0],
[3.0,5.0,6.0]])
x = np.array([1.0,2.0,3.0])
# derivative of matrix vector product with respect to vector
deriv_matvec_vec = ta.diff(matvec_vec,x,1,args=(A,))
print(f'df / dx :\n{deriv_matvec_vec}\n')
# derivative of matrix vector product with respect to matrix
deriv_matvec_mat = ta.diff(matvec_mat,A,1,args=(x,))
print(f'df / dA :\n{deriv_matvec_mat}')
Namesake
The package is named after Brook Taylor, the namesake for Taylor series and the originator of finite differences.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distributions
Built Distribution
File details
Details for the file taylor-0.0.0-py3-none-any.whl
.
File metadata
- Download URL: taylor-0.0.0-py3-none-any.whl
- Upload date:
- Size: 7.0 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/0.0.0 importlib_metadata/4.10.0 pkginfo/1.8.2 requests/2.26.0 requests-toolbelt/0.9.1 tqdm/4.62.3 CPython/3.9.9
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | eaaba7e3c76a0b01fb642575a333c4b27e41624668ee2682d5f92ad978104d2f |
|
MD5 | 2261e907e019fa74cb0509a185e8af05 |
|
BLAKE2b-256 | 37876e5dabb96763b93eaf3e94d8258ecc79d87c7a3becba47db899b4a9c6451 |