Generate all possible distributions for nonlinear optimization problem

I have set out an optimization problem in hope to find a new distribution for my project. However, the equations are rather complex and I am not sure whether the built-in matlab functions can do the job. So, I want to try simple trial and error by generating all the possible distributions first and then testing it one by one. Can someone help me generate them please? Max degree : 100; Possible degrees range from : 0:0.0001:1; Sum of degrees must be between: 0.9 - 1.1 (ideally 1)

Initially, I tried to generate them by using permutation with repetition however the size of such a matrix would have (10^4)^100 rows ( set to INF in MATLAB).

Any suggestions would be much appreciated!

Answer 1

[Jan]

Wow. You ask for a method to create 10^400 numbers, although you are aware of the problem, that this will not fit into the RAM or the hard disk of your computer. The sum-criterion reduces the number, but as far as I can see not dramatically enough. While the universe has about 10^80 protons only you would even need a very efficient compression method to store these data anywhere. A problem would be the extremely high information density which is required, if the data should be processed in finite time. This density could lead to the creation of a black hole. If so, please ask your local security administration, if this is covered by your employment agreement.

Of course this means, that you cannot apply a brute force method. A genetic algorithm will be more efficient, e.g. create fairly good results in finite time, perhaps hours.

[MLUtility767]

I've got a similar problem, how do I set up a permutation optimization with GA? My fitness function has 12 variables that can accept 20 different numbers each. The problem is that there must be not 2 equal numbers inside the function. I.E: if var1 gets 5 as a random generated number by first population the second variable can get every 20 number except 5... and so on.