Variable versus No. of Indices?!
Mike
A decision-variable is defined as follows: VarA(i1, i2, i3, ...iN)
When a variable is addressed in a mathematical programming model, are
there rules on what constitutes a wise number of indices and which
ones?
Of course, the higher the value of N, the more memory the solve needs.
Say i3 is i2's attribute, shouldn't in that case be better to exclude
i3,....and so on?
Can "Total Independency" among variable's indices be the rule for
selection? So, if i3 is an attribute to i2, then i3 better be excluded
from being variable's index and instead gets defined as an attribute
to i2
Cheers,
Mike
Paul
Can you clarify what you mean by an index being an attribute of
another index?
/Paul
Mike
Suppose a math model MP has a major variable named: Car_Var(Model,
Year, Warranty, Price, Mileage, NDoors) has 6 indices as shown where
each might have few ones to few 100s of elements.
So, the less the number of "sets" or "indices" the better when it
comes to memory. To reduce the number from 6 sets to, say, 5, one
might remove the "NDoors" index and create a table that links "Model"
to "NDoors" for example since NDoors is an attribute to car Model!
Thanks,
Mike
Ray Vickson
Your illustration is not convincing. For example, if there are two
models and each can have two or four doors, there are a total of 4
combinations of model/door number, irrespective of whether or not you
keep the indices separate or combine them. You would save, however, if
some models come in only one door number. Anyway, smart MP modelling
languages will not construct more variables than needed, so would
construct, say, only X(model1_4door), X(model1_2door) and X
(model2_4door) even if regarded as X(model,doors), provided that you
have somewhere else given the allowable (model,door) combinations to
exclude model2 and 2door.
R.G. Vickson
>
> Thanks,
> Mike