Hi,

I am currently trying to generate multiple random numbers in C for different variable:-


die1=1+(rand()%5);
die2=1+(rand()%5);
die3=1+(rand()%5);
die4=1+(rand()%5);

But I need to contrain the total of die1, die2,die3 and die4 to be 5 as well. If i insert die1+die2+die3+die4=5, i do not think it will constrain the distribution.


Please advise. Thanks.


-Jason

Your four variables are integers, each set to 1 + some integer. Three of them must stay at 1 and the fourth one can be two. Pick a number between 1 and 4 and increment only one variable based on that. You pick a 3, die3=2 and the rest stay at 1.

Yeah. I understood the concept. But do we have ways to automate them to generate all sorts of combination in this constrain?

I tried to find from internet related to these cases; they arent related. Most of them are just giving random numbers generation.

As I mentioned previously, if i constrain the total value after the random number is generated for these variables, it wont work because it is not dependent. Unless I can slot something in between the rand?

Thanks.

-Jason

If you had, say, 100, to distribute 5 into 5 variables: You pick a number from 1 to 100, let's say you get 20. The first variable gets the 20. Now 100 - 20 = 80 is left. So pick a number between 1 and 80. You stop when you get to only 1 variable and it gets the remainder. Or if the amount left reaches zero, you stop early and the rest of the variable must be zero.

Hi,

I get the idea now. Thanks. But i just have a thought if that is the case, would it be very bias to have the top variables always get the most of the values rather than the bottom ones?

What I expected at the first place is to have fair distribution of cake among the variables.

let say i have only consider 3 possbile values, (0,1 and 2). And have 5 variables. If I did consider the below idea, in all the cases, the second of third variables always get the most rather than the fourth and fifth variable.

Please advise. Thanks..

Rgrds,
Jason

I can't help you there since your test case is so very constrained. Except to say that only a very few possible outcomes are going to satify you. Enumerate them and pick one.

Getting back to my test case, you could preallocate 10 to each variable and then distibute the remaining 50 as I indicated. Or you could distribute the numbers one at a time. Cycle though each variable and generate a 0 or 1. Add that to each variable and subtract it from the mass to be distributed.

jason Perderabo idea is cool. but if u want ot " distribute piece of cake evenly" then you can write an additional code

n=1+rand()%4
switch(n)
{
case 1:
// use perderabo idea starting with die1 - die4

case 2:

// use perderabo idea starting with die2 - die4,die1 in circular way

case 3:
//and so on...........
}

i know its not a very pretty solution but considering your unique constraints this is the best solution i got