Kevin,
sorry I don't understand this sentence:
"It has been suggested that I do a cluster analysis. Wouldn't this bet
mepart way there?"
Let your products be numbered 1, ..., n, and let p(i) be the location
where product i is stored (assigned to) in the warehouse. Then p is a
permutation of th
Just in case you are still interested in theoretical aspects:
In combinatorial optimization, the problem you describe is known as the
Quadratic (Sum) Assignment Problem (QAP or QSAP) and is well known to
arise in facility and warehouse layouts. The task itself is considered
hard, comparable to th
I have decided to use this SANN approach to my problem but to keep the run time
reasonable instead of 20,000 variables I will randomly sample this space to get
the number of variables under 100. But I want to do this a number of times. Is
there someone who could help me set up WINBUGS to repeat
> Ph: (410) 502-2619
>
> Fax: (410) 614-9625
>
> Email: rvarad...@jhmi.edu
>
> Webpage:
> http://www.jhsph.edu/agingandhealth/People/Faculty/Varadhan.html
>
>
>
>
>
>
>
> -Original Message-----
> From: r-help-b
On Thu, Apr 2, 2009 at 3:49 PM, wrote:
> Sorry I sent a description of the function I was trying to minimize but I
> must not have sent it to this group (and you). Hopefully with this clearer
> description of my problem you might have some suggestions.
>
> It is basically a warehouse placement
lf Of Florin Maican
Sent: Thursday, April 02, 2009 11:33 AM
To: r-help@r-project.org
Subject: Re: [R] Constrined dependent optimization.
I tried many optimizers in R on my large scale optimization problems.
I am not satisfied with their speed on large op problems. But you may try in
this order
Just a quick thought: if you have the locations
of the items along a line segment (the "warehouse"),
can you find a way to formulate the problem so that
the average distances represent a matrix calculation
(i.e., multiply some huge matrix by the locations)
then it's a linear problem ...
It's t
I tried many optimizers in R on my large scale optimization problems.
I am not satisfied with their speed on large op problems. But you may
try in this order
nlminb
ucminfucminf package
spq BB package
optim
Is here someone that try to port Ipopt in R?
https://projects.co
Sorry I sent a description of the function I was trying to minimize but I must
not have sent it to this group (and you). Hopefully with this clearer
description of my problem you might have some suggestions.
It is basically a warehouse placement problem. You have a warehouse that has
many item
As I told you before, without knowing the definition of your function
f, one cannot help much.
Paul
On Wed, Apr 1, 2009 at 3:15 PM, wrote:
> Thank you I had not considered using "gradient" in this fashion. Now as an
> add on question. You (an others) have suggested using SANN. Does your answe
rkevinburton wrote:
>
> Thank you I had not considered using "gradient" in this fashion. Now as an
> add on question. You (an others) have suggested using SANN. Does your
> answer change if instead of 100 "variables" or bins there are 20,000? From
> the documentation L-BFGS-B is designed for a
Thank you I had not considered using "gradient" in this fashion. Now as an add
on question. You (an others) have suggested using SANN. Does your answer change
if instead of 100 "variables" or bins there are 20,000? From the documentation
L-BFGS-B is designed for a large number of variables. But
Apparently, the convergence is faster if one uses this new swap function:
swapfun <- function(x,N=100) {
loc <- c(sample(1:(N/2),size=1,replace=FALSE),sample((N/2):100,1))
tmp <- x[loc[1]]
x[loc[1]] <- x[loc[2]]
x[loc[2]] <- tmp
x
}
It seems that within 20 millions of iterations, one gets th
Optim with SANN also solves your example:
---
f <- function(x) sum(c(1:50,50:1)*x)
swapfun <- function(x,N=100) {
loc <- sample(N,size=2,replace=FALSE)
tmp <- x[loc[1]]
x[loc[1]] <- x[loc[2]]
x[loc[2]] <- tmp
x
}
N <- 100
opt1 <-
optim(fn=f,par=sam
Actually, one can use lpSolve to find a solution to your example. To
be more precise, it would be necessary to solve a sequence of linear
*integer* programs. The first one would be:
max f(x)
subject to
x >= 0
x <= 100
sum(x) = 100.
>From this, one would learn the optimal position of the number
Image you want to minimize the following linear function
f <- function(x) sum( c(1:50, 50:1) * x / (50*51) )
on the set of all permutations of the numbers 1,..., 100.
I wonder how will you do that with lpSolve? I would simply order
the coefficients and then sort the numbers 1,...,100 accord
rkevinbur...@charter.net wrote:
> I am sorry but I don't see the connection. with SANN and say 3
> variables one of the steps may increment x[1] by 0.1. Not only is
> this a non-discrete integer value but even if I could coerce SANN to
> only return discrete integer values for each step in the opti
I do not really understand your argument regarding the non-linearity
of f. Perhaps, it would help us a lot if you defined concretely your
objective function or gave us a minimal example fully detailed and
defined.
Paul
On Mon, Mar 30, 2009 at 1:16 PM, wrote:
> It would in the stictess sense be
It would in the stictess sense be non-linear since it is only defined for
descrete interface values for each variable. And in general it would be
non-linear anyway. If I only have three variables which can take on values
1,2,3 then f(1,2,3) could equal 0 and f(2,1,3) could equal 10.
Thank you f
On Sun, Mar 29, 2009 at 9:45 PM, wrote:
> I have an optimization question that I was hoping to get some suggestions on
> how best to go about sovling it. I would think there is probably a package
> that addresses this problem.
>
> This is an ordering optimzation problem. Best to describe it wit
rkevinburton wrote:
>
> I have an optimization question that I was hoping to get some suggestions
> on how best to go about sovling it. I would think there is probably a
> package that addresses this problem.
>
> This is an ordering optimzation problem. Best to describe it with a simple
> exam
I have an optimization question that I was hoping to get some suggestions on
how best to go about sovling it. I would think there is probably a package that
addresses this problem.
This is an ordering optimzation problem. Best to describe it with a simple
example. Say I have 100 "bins" each wit
22 matches
Mail list logo