search

rahlk / Experimental-Algorithms

CSC 591: Experimental Algorithms
http://courses.ncsu.edu/csc591/lec/003/
MIT License
0 stars 0 forks source link

Parallelized GALE #6

Open bigfatnoob opened 8 years ago

rahlk commented 8 years ago

@bigfatnoob Can you describe you're parallelization strategy, also include other strategies that you think might work.

bigfatnoob commented 8 years ago

Yeah will do that

rahlk commented 8 years ago

Reply to this thread. Also, is this your strategy — http://charm.cs.uiuc.edu/research/masterSlave?

rahlk commented 8 years ago

Or, this?

Island Model: In the "island" approach to parallelization of genetic programming, the population for a given run is divided into semi-isolated subpopulations (called demes). Each subpopulation is assigned to a separate processor of the parallel computing system. The run begins with the one-time random creation of a separate population of individuals at each processor of the parallel computer system. A typical run of genetic programming on today's computers might entail randomly creating between perhaps 500 and 30,000 individuals at each processor. This process of initial random creation takes place in parallel at each processor. As soon as each separate processor finishes this one-time task, it begins the main generational loop.

bigfatnoob commented 8 years ago

Yeah this perfectly describes what I have done. Coincidence. I havent read this. Will include it

rahlk commented 8 years ago

Island Model? Or MasterSlave?

bigfatnoob commented 8 years ago

No its not master slave as of now. I will try out master slave as it requires more changes in the implementation

rahlk commented 8 years ago

Sure.. I'm making changes offline, you can just paste the parts here. I'll incorporate it.

bigfatnoob commented 8 years ago

Cool. Thanks man