[Previous] [Next]

SLPDATA("SB")

---

Purpose

Assign an initial value for the step bound of a variable

Synopsis

procedure SLPDATA(SB:string, Var:mpvar, Value:real)

Arguments

SB	string containing "SB" indicating the SLPDATA type
Var	the variable being assigned an initial step bound
Value	initial value to be assigned to the step bound

Example

The following code fragment assigns initial step bounds of 1.2 to all the variables RVP(P):

declarations
  RVP, RVI:array(Prod) of mpvar
  RVPq:array(Prod) of genctr
end-declarations

forall (P in Prod) RVPq(P):= RVP(P) = RVI(i)^1.25
forall (P in Prod) SLPDATA("SB",RVP(P),1.2)

Further information

Step bounds are used in SLP optimization to restrict the amount by which the value of a variable can change between successive SLP iterations. A smaller value means that the linearization is more accurate but that it takes more iterations to change a value by a large amount; a larger value means that it is easier for the value of the variable to change but the linear approximations are less accurate.

The SLP optimizer will estimate step bounds for all variables once optimization has begun. If step bounds are required from the beginning, or if the estimates are not appropriate, then initial values should be provided.

[Previous] [Next] [Index]

---

If you have any comments or suggestions about these pages, please send mail to docs@dashoptimization.com.