Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Re: numerical gradient, Jacobian, Hessian: missing from Mathematica?

229 views
Skip to first unread message

Bob Hanlon

unread,
Apr 21, 2014, 11:00:11 PM4/21/14
to

$Version

9.0 for Mac OS X x86 (64-bit) (January 24, 2013)

n=2;

vars=Table[x[m],{m,n}]

{x[1],x[2]}

funcs=Table[f[m]@@vars,{m,n}]

{f[1][x[1],x[2]],f[2][x[1],x[2]]}

http://reference.wolfram.com/mathematica/ref/Grad.html

Grad[funcs, vars]

{{Derivative[1, 0][f[1]][x[1], x[2]], Derivative[0, 1][f[1]][x[1], x[2]]},
{Derivative[1, 0][f[2]][x[1], x[2]], Derivative[0, 1][f[2]][x[1], x[2]]}}

Grad is new in v9; for some older versions use D[funcs, {vars}]

Grad[funcs, vars]==D[funcs, {vars}]

True

http://mathworld.wolfram.com/Jacobian.html

jacobian=Outer[D,funcs,vars]

{{Derivative[1, 0][f[1]][x[1], x[2]], Derivative[0, 1][f[1]][x[1], x[2]]},
{Derivative[1, 0][f[2]][x[1], x[2]], Derivative[0, 1][f[2]][x[1], x[2]]}}

Det[jacobian]

Derivative[0, 1][f[2]][x[1], x[2]]*Derivative[1, 0][f[1]][x[1], x[2]] -
Derivative[0, 1][f[1]][x[1], x[2]]*Derivative[1, 0][f[2]][x[1], x[2]]

http://mathworld.wolfram.com/Hessian.html

hessians=D[#,{vars,2}]&/@funcs

{{{Derivative[2, 0][f[1]][x[1], x[2]], Derivative[1, 1][f[1]][x[1], x[2]]},
{Derivative[1, 1][f[1]][x[1], x[2]],
Derivative[0, 2][f[1]][x[1], x[2]]}},
{{Derivative[2, 0][f[2]][x[1], x[2]],
Derivative[1, 1][f[2]][x[1], x[2]]},
{Derivative[1, 1][f[2]][x[1], x[2]],
Derivative[0, 2][f[2]][x[1], x[2]]}}}

Det/@hessians

{-Derivative[1, 1][f[1]][x[1], x[2]]^2 + Derivative[0, 2][f[1]][x[1], x[2]]*
Derivative[2, 0][f[1]][x[1],
x[2]], -Derivative[1, 1][f[2]][x[1], x[2]]^2 +
Derivative[0, 2][f[2]][x[1], x[2]]*Derivative[2, 0][f[2]][x[1], x[2]]}


Bob Hanlon



On Mon, Apr 21, 2014 at 5:11 AM, Alan <alan....@gmail.com> wrote:

> Am I overlooking commands to produce numerical gradient, Jacobian, and
> Hessian? The closest I could find was ND (in the NumericalCalculus
> package), which seems to be only for univariate functions.
>
> Thanks,
> Alan Isaac
>
>

Alan

unread,
Apr 25, 2014, 12:44:27 AM4/25/14
to
Hi Bob, When symbolic differentiation is appropriate, `D` (or `Grad`) meets my needs. But in this case I'm looking for numerical gradient, Jacobian, and Hessian. Thanks, Alan

Alexey Popkov

unread,
Apr 27, 2014, 2:07:03 AM4/27/14
to
For numerical gradient, Jacobian and Hessian you could use undocumented Experimental`NumericalFunction, see here:

http://mathematica.stackexchange.com/a/44890/280

0 new messages