Genetic algo shows random results, is it ok ?

PanagiotisCharalampous
16 Oct 2017, 11:47

Hi goracioo,

Yes this is a normal behavior. It is the way the genetic algorithm works. The genetic algorithm starts searching at a random point i.e. by selecting a random set of parameters, and looks for local maximum. If there is no improvement in new population (new sets of parameters), it will be stopped assuming that it reached a local maximum. In contrast, the grid method tries to find the global maximum by checking all possible parameters combinations. So the grid method is an exhaustive search and will always reach to the same conclusion while the genetic algorithm might give different results each time.

Best Regards,

Panagiotis

goracioo
16 Oct 2017, 13:15

RE:

Panagiotis Charalampous said:

Hi goracioo,

Yes this is a normal behavior. It is the way the genetic algorithm works. The genetic algorithm starts searching at a random point i.e. by selecting a random set of parameters, and looks for local maximum. If there is no improvement in new population (new sets of parameters), it will be stopped assuming that it reached a local maximum. In contrast, the grid method tries to find the global maximum by checking all possible parameters combinations. So the grid method is an exhaustive search and will always reach to the same conclusion while the genetic algorithm might give different results each time.

Best Regards,

Panagiotis

 

Ok, thanks. About Grid search it's very heavy right now. I did some tests with grid and eats machine reasources like crazy. For example 200k passes eats about 3 gigs of memory. MT5 solved this very efficiently as they load workers or agents. So if you have 2 cores you have 2 working agents. Also you can buy agents from MQL5 network and do pretty complex test in minutes. Hope next evolution of cAlgo would be something like this.