Question

i am planning to use tinyGP as a way to train a set of Input variables (Around 400 or so) to a value set before. Is there a maximum size of Input variables? Do i need to specify the same amount of variables each time?

I have a lot of computation power (500 core cluster for a weekend) so any thoughts on what parameters to use for such a large problem?

cheers

Was it helpful?

Solution

In TinyGP your constant and variable pool share the same space. The total of these two spaces cannot exceede FSET_START, which is essentially the opcode of your first operator. By default is 110. So your 400 is already over this. This should be just a matter of increasing the opcode of the first instruction up to make enough space. You will also want to make sure you still have a big enough "constant pool".

You can see this checked with the following line in TinyGP:

if (varnumber + randomnumber >= FSET_START ) 
   System.out.println("too many variables and constants");
Licensed under: CC-BY-SA with attribution
Not affiliated with StackOverflow
scroll top