[Date Index] [Thread Index] [Author Index]

returning inverse Hessian from FindMinimum

• To: mathgroup at smc.vnet.net
• Subject: [mg87950] returning inverse Hessian from FindMinimum
• From: samelfors at gmail.com
• Date: Mon, 21 Apr 2008 03:24:43 -0400 (EDT)

I am using Method -> "QuasiNewton" with FindMinimum. I specify the
Gradient but not the Hessian. Can I have FindMinimum return the
inverse Hessian it has accumulated when it completes its
optimization?

The implementation notes indicated BFGS is used for QuasiNewton. I
would like to use the inverse Hessian to get an idea of how stable the
fitted parameters are around the minimum. I would also like to use it
to fudge together the minima from running FindMinimum[] in batch
across several machines using the parallel computing toolkit.

I don't think this is possible or desirable - I'm just messing around.
I looked through the docs and the source for
Optimization`UnconstrainedProblems` and didn't find anything, well,
other than some good programming pointers.

Any suggestions or general guidance would be cool.

Cheers,
Sam