The genetic algorithm and its possible applications - page 2

New comment
 
The target is set in the built-in optimiser, just not directly.

1. you set your indicator as a function defining the strategy profit curve
2. the optimizer takes the obtained profit of the strategy as its target

Possible applications of GA are mentioned here
 
If I understand correctly, "interbreeding" can be done between identical "genes" (parameters) of individuals (copies of the same system with different values of the same parameters).

In this case, the "crossing" should be a simple average (sum of two values divided by two, of two parameters, of two identical in structure "chromosomes")?

I mean, the values of opposing parameters of copies of one system are added up, divided by two, and a third copy of the system is obtained, with the average values of the parameters?
 
RomFil:
there is no way to make your target

OnTester() doesn't work?

 
Реter Konow:
If I understand correctly, "interbreeding" can be done between identical "genes" (parameters) of individuals (copies of the same system with different values of the same parameters).

In this case, should the 'crossing' be a simple average (the sum of two values, two parameters, two identical 'chromosomes' in structure)?

You can do whatever you like, there are dozens of variations.

 
Dmitry Fedoseev:

https://www.mql5.com/ru/search#!keyword=proprietary%20criterion%20optimization

Great! I didn't know five had such a feature ... :) I'm still working on four... :) Thanks.

That's it then. The question is off ... :)
 
RomFil:

Great! I didn't know five had such a feature ... :) I'm still working on four... :) Thanks.

That's it then. The question is off ... :)
Actually, "this feature" has been around since four ))))
 
SeriousRacoon:
Actually, "this feature" has been around since four ))))
So I'm a "nerd" ... :)
 
Реter Konow:
The approach itself is not enough. It can serve as a tool for 'polishing' finished systems - testing and debugging.
The systems themselves can't be built like this. You need a different algorithm. When it is invented, GA will be remembered.

I've seen such methods from guys who work on so-called "artificial life" projects. Interesting stuff. At first glance, it looks like mush, but genetics makes "intelligent" genes out of mush through millions of passes.

 
Aleksey Mavrin:

I've seen these methods from guys who work on so-called "artificial life" projects. Interesting stuff. At first glance, it looks like mush, but genetics makes intelligent genes out of mush through millions of passes.

You mean "smart" - optimised and adapted to a certain environment?

OK. We create a closed environment, from a complex of interacting systems. For example: three-gene individuals "devour" (attach and grow) two-gene individuals if they cannot "escape" from them (let us say genes X and Y are coordinates on the plane). The three-gene individual has a gene Z added - the third dimension. It has an advantage. The task of the two-gene individual is to survive and not be caught, and the task of the three-gene individual is to catch the two-gene as quickly as possible and save time and resources.

The goal: the genetically weaker individual must calculate the genetically stronger one to survive.
 
Aleksey Nikolayev:

OnTester() is not suitable?

Yes, in the genetic optimization mode by "Maximum of user criterion" the GA of the tester will maximize the result obtained from double OnTester() - everything works, but there is a problem in the automation "to perk up the GA" when it starts to converge around the found local maximum and does not want to search for other options by input parameters, in general, in the fine tuning you have to help GA by adding conditions to OnTester() or by splitting optimization parameters into several intervals

123456789...15
New comment
Reason: