Re: [Help-glpk] Can't use variable within abs()?

[glpkuser]

Andrew Makhorin wrote:
> 
>> I'm getting an error saying that "argument for abs has invalid type".
> 
>> I'm trying to model a load-balancing scenario where 10 weighted
>> items are distributed across 3 buckets. I want to minimise the
>> difference between each bucket's total weight and the average weight
>> of all buckets, so I'm trying to do the following, which is not
>> working:
> 
>> minimize z:  sum{b in BUCKETS} abs(BucketWeight[b] - AveBucketWeight);
> 
> You cannot use variables in abs(), because this leads to non-linear
> objective not allowed in glpk.
> 
> However, in your case the objective is piecewise linear and convex,
> so it can be easily reformulated as follows:
> 
> var u{b in BUCKETS}; /* auxiliary variable */
> 
> s.t. aaa{b in BUCKETS}: u[b] >= BucketWeight[b] - AveBucketWeight;
> 
> s.t. bbb{b in BUCKETS}: u[b] >= AveBucketWeight - BucketWeight[b];
> 
> minimize z: sum{b in BUCKETS} u[b];
> 
> 


Thank you for your help. I'm a beginner with linear programming and O.R.;
could you point me to a reference where I can find more info on piecewise
linearity and convex-ness with respect to re-formulating or fixing
non-linear models? For example, would the Winston book on Operations
Research have it?
 
-- 
View this message in context: 
http://www.nabble.com/Can%27t-use-variable-within-abs%28%29--tf4338967.html#a12360372
Sent from the Gnu - GLPK - Help mailing list archive at Nabble.com.