Originally posted by: dirkfluss

Hi, there

i am a rookie working on CPLEX and encountered a problem withe square root.

I was trying to build up a model in CPLEX and want to set a contrain that contains a square root term. I am not sure if this is not allowed in a linear programming. The idea was to ensure a sum of several Gamma distributed variables will not excess a certain amount of capacity. I used the Central Limit Theorem to transform it in an approximated normal distribution. Then a have the constraint like followed:

sqrt( sum( k in P) ( alphar[k] *q[k] )) *beta * Z + sum (k in P) (alphar[k] *q[k] ) * beta <= K

But the Software connot extract this constraint.

It will be very kind if you can help me this out; thanks !

Dirk

Originally posted by: fbahr

CPLEX requires that constraints define convex regions.

Assuming that the q[k]s are your decision variables: sqrt of [a function of] q[k] is concave, though.

--fbahr

Originally posted by: dirkfluss

oh i see. 

Thanks for replying

Dirk