`scipy.optimize._minpack:_chkder`

Signature

built-in_chkder(m, n, x, fvec, fjac, ldfjac, xp, fvecp, mode, err)

Summary

Check the gradients of a function against finite differences.

Extended Summary

Utility function for verifying correctness of user-supplied Jacobians.

Parameters

m : int: Number of functions (residuals).
n : int: Number of variables.
x : ndarray, shape (n,): Point at which to check derivatives, float64.
fvec : ndarray, shape (m,): Function values at x, float64. Used only when mode=2.
fjac : ndarray, shape (m, n): User-supplied Jacobian at x, float64. Used only when mode=2.
ldfjac : int: Leading dimension of fjac.
xp : ndarray, shape (n,): Perturbed point for finite differences, float64. Modified when mode=1, used when mode=2.
fvecp : ndarray, shape (m,): Function values at xp, float64. Used only when mode=2.
mode : int: 1: initialize xp for gradient checking. 2: compute error between analytical and finite-difference gradients.
err : ndarray, shape (m,): Error estimates for each function, float64. Modified when mode=2.

Returns

: None: Modifies xp (when mode=1) or err (when mode=2) in-place.

Notes

This function is called twice: 1. mode=1: Computes perturbed point xp from x. 2. mode=2: Compares analytical Jacobian fjac with finite differences and stores relative errors in err.

Aliases

scipy.optimize._minpack._chkder