The genetic algorithm and its possible applications - page 5
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
1. it's too much, the graph is scaled so that you can't see the useful results. I return a value slightly higher than the worst Custom. The main thing, though, is to set the right direction for improvement.
2. what's the point? The main thing is to set the right direction, which means you have to show GA that it showed the worst result here, not just a weak one.
1. this is a disadvantage of the regular optimizer display, but does not mean that you should actually consider the lack of research tool (in this case, the optimizer MT), in order to get a better result. in fact, the right approach to visualize the optimization results yourself, because the optimizer MT does not know what you really need. There is no tool for showing results in the optimization table (and on the graph) at the moment, I think it will be implemented one day.
2. No, the main thing is not just to show that the result is "bad", but to show that the result is "very bad", it makes a huge difference for AO.
I don't remember if I wrote about it on the forum, but it's really a problem and it's not clear why it's implemented in MT. In theory, if the examiner returned an error code "wrong parameters", the tester is obliged to generate another instance instead, so that the population is complete.
Absolutely agree.
Maybe for future generations (except first) such trick won't work (question for GA experts), but for first sample (which is random anyway) replacement of one random set (with wrong inputs) by another won't do any harm. And the probability of encountering incorrect parameters in future generations will be much lower. Strange that they won't...
Tell me, are there any developments in GA for a variable number of parameters?
Software implementation is not a problem. The same sets of pairs are "crossed" in MT. It is possible to implement "mutation", then the sets can be arbitrary.
Absolutely agree.
Maybe for future generations (except first) such trick won't work (question for GA scholars), but for first sample (which is random anyway) replacement of one random set (with wrong inputs) by another won't do any harm. And the probability of encountering incorrect parameters in future generations will be much lower. Strange that they won't...
No obstacle, generation is always a sampling from a pool of possible combinations of pairs, only if the pool is not enough, but even then something can be invented, clones for example.
No obstacle, generation is always a selection from a pool of possible combinations of pairs, only if the pool is insufficient, but even then you can think of something, clones for example.
Why don't they do it then? They're not idiots.
Absolutely agree.
Maybe for future generations (except first) such trick won't work (question for GA experts), but for first sample (which is random anyway), replacement of one random set (with wrong inputs) with another won't do any harm. And the probability of encountering incorrect parameters in future generations will be much lower. Strange that they won't.
Good - incorrect input parameters variant should be ignored by optimizer and instead of it another one should be generated, so that population would be always full. If the number of possible variants is insufficient - duplicates taken with probability proportional to the rank of an individual in the population are acceptable.
Tell me, are there any GA developments for a variable number of parameters?
I think it is unlikely. Based on the concept of GA, identical structures - clones of the same system with different parameter values - can interbreed. Different individuals within the same environment cannot interbreed in Nature either. This natural biological constraint stops the emergence of ridiculous, non-viable freaks that are meaningless to the ecosystem. Such "experiments" always end in failure and are only suitable for laboratory research. GA imitates biology and therefore does not deviate from the principles of interbreeding, inheritance and selection.
The question, from a theoretical point of view, is very interesting. Evolution creates not only "optimised" versions of creatures in the course of their "adjustment" to conditions, but also fundamentally new species. Where do they come from if interspecies interbreeding is impossible? So they come from natural mutations. But, - mutation is a change in existing genes, not acquisition of new ones. That is, - the set cannot be increased, and "calibration" only adapts (optimizes) the living species. Where do new and more complex creatures come from?
Even if we make an algorithm randomly "cast" parameters into arbitrary systems and also randomly find an optimization target (fitness function) for them, what can this give us?
Tell me, are there any developments in GA for a variable number of parameters?
is.
The practical applications are quite extensive - from genetic programming to calculating the shape and volume of bodies, taking into account strength maximisation and volume minimisation.
there is.