[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: help installing functions from repository (was: Re: Built-in paralle
From: |
Olaf Till |
Subject: |
Re: help installing functions from repository (was: Re: Built-in parallelization in optimization functions) |
Date: |
Mon, 23 Jun 2014 14:32:31 +0200 |
User-agent: |
Mutt/1.5.21 (2010-09-15) |
Hi Michael,
On Mon, Jun 23, 2014 at 04:53:09AM -0700, Michael Creel wrote:
> I have played around in the past with some ideas for using a battery of SA
> minimizers. If I recall correctly, this gave some improvements in getting to
> the neighborhood of the global minimum, during the early part of the
> iterations, but convergence in the later part was no faster than the public
> version.
maybe you can specify the approach ... ? But surely the resulting
algorithm actually won't be real simulated annealing anymore ...
> An obvious way to do some parallelization in optim would be to do numeric
> gradients and/or Hessians in parallel. That would be a generic, general
> purpose improvement.
For gradients, this is already done (in the native numeric gradient
functions of the new frontends). As for Hessians, algorithms needing a
Hessian seem to use BFGS, usually ...
Olaf
--
public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net
signature.asc
Description: Digital signature