verry.autodiff.grad

verry.autodiff.grad(fun)

Return a function that evaluates the gradient of the multivariate scalar-valued function.

Parameters:

fun (Callable) – Diffarentiated function.

Returns:

Gradient of fun.

Return type:

Callable

Warning

fun must neither be a constant nor contain conditional branches (cf. Common pitfalls).

Examples

This example computes a value of the gradient.

>>> from verry import FloatInterval as FI
>>> from verry import function as vrf
>>> f = lambda x, y: vrf.sqrt(x * y + 3)
>>> df = grad(f)
>>> c0 = df(0.5, 1.0)
>>> print(format(c0[0], ".6g"), format(c0[1], ".6g"))
0.267261 0.133631

To obtain the value that accounts for rounding errors, pass intervals to df.

>>> c1 = df(FI("0.5"), FI("1"))
>>> print(format(c1[0], ".6g"))
[0.267261, 0.267262]