Decision Optimization

Originally posted by: victortkl

hi, how can i create a variable that is in normal distribution. in the reference, only rand() and rand(int) is available.
#DecisionOptimization
#OPLusingCPLEXOptimizer

Hi,

have you thought about using Box Muller transform?

I would write

int n=1000000;
 int BIG=1000000;
 
 float x[i in 1..n]=-1+2*rand(BIG)/BIG;
 float y[i in 1..n]=-1+2*rand(BIG)/BIG;;
 
 float w[i in 1..n]=sqrt(x[i]*x[i]+y[i]*y[i]);
 
 
 float w2[i in 1..n]=(w[i]<=1)?(-2*ln(w[i])/w[i]):-1;
 
 float x2[i in 1..n]=(w[i]<=1)?x[i]*w2[i]:-1;
 float y2[i in 1..n]=(w[i]<=1)?y[i]*w2[i]:-1;

the values of x2 and y2 when w<=1 obey a normal distribution

Regards

Alex
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: victortkl

thank you. it's a good idea
but it needs much codes.
now i prefer to generate the random value in database...
#DecisionOptimization
#OPLusingCPLEXOptimizer

Hi,

what you can also do if you accept to use external ways is to use the java class java.util.Random with the function nextGaussian.
You are able to call this from OPL with the function IloOplCallJava

regards

Alex
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: victortkl

thank you. but using java is a bit difficult to me. :-)
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: Ferenc Katai

Hello there,
may I know in what context do you want to use normal distribution? is a search you are designing or maybe some perturbation to data?
cheers
#DecisionOptimization
#OPLusingCPLEXOptimizer

Hi,

full example at https://github.com/AlexFleischerParis/howtowithopl/blob/master/randomnumber.mod

And you can also rely on java rand through IloOplCallJava

regards

------------------------------
[Alex] [Fleischer]
[Data and AI Technical Sales]
[IBM]
------------------------------

Original Message:
Sent: Mon June 20, 2022 06:39 AM
From: Archive User
Subject: Re: random value in normal distribution

Originally posted by: Ferenc Katai

Hello there,
may I know in what context do you want to use normal distribution? is a search you are designing or maybe some perturbation to data?
cheers
#DecisionOptimization