
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
What criteria and how many offspring should be killed . How many parents and descendants should be depending on the number of optimised parameters .From what epoch incest can be allowed . In general it is not quite clear on what principle to kill parents .Just unfortunately your labour is not quite suitable for my purposes , but so thank you very much with a brush .
You don't have to kill anyone. It can work even with one chromosome (crossbreed with itself). And so on average 50 chromosomes are enough for all purposes.
You just don't understand the algorithm.
The algorithm is universal, suitable for any purpose, that's what it's called UGA. There are just some tasks that can be solved faster in other ways.
1. Give me a hint plz....
2. by what criterion and how many descendants to kill .
3. How many parents and descendants should be depending on the number of parameters to be optimised .
4. From which epoch incest can be allowed .
5. In general it is not quite clear on what principle to kill parents .
6. Just unfortunately your labour is not quite suitable for my purposes , but so many thanks with a brush .
1. I am not quite clear on the purpose of your post. If you need advice on implementing your algorithm, you need details, a general concept, so to speak, so that there would be something to discuss.
If the questions are about the algorithm described in the article, see 2, 3, 4, 5, 5. 2, 3, 4, 5, 6.
2. Descendants are not killed. Duplicates of parents and their descendants are killed.
3. 50 individuals in the bulk of the population is the most commonly used meaning.
4. Interbreeding "with itself" is not allowed. But if it is very necessary (the population is "dying out") - it is possible. :) several attempts are made to find a suitable "partner".
5. Parents are not "killed" randomly or under some conditions, but are replaced by offspring (exactly half of the population is replaced by offspring - if, of course, the population is full).
6. If the problem is very "specific" - perhaps (if it is possible at all) try to find an analytical solution first.
Thank you for your prompt response. I need genetics for various experiments. Your biblioteka as it seems to me is not very convenient in use so I am writing my own. Unfortunately in the network poorly described the whole algorithm mainly only GA determiners, if you give a link to the description (in Russian) I will be very grateful. Below I will describe how I see it all, if not difficult to correct .
1 Randomly create parental individuals, not less than 50.
2 Generating from them by crossbreeding and mutations colony of offspring number not less than ..... nowhere found a description.
3 Looking for duplicates, the parental one is removed.
5 If necessary, completes colonies of descendants and parents. Parents are created randomly descendants naturally .
4 Run all of them through the FF. The FF is placed outside the GA.
5 We rank them all. The best one is given the status of Rambo. The higher the ranking the more likely to become a daddy.
6 Divide the whole crowd of individuals into strong and weak .The strong ones will be parents at the next step.
7 Cut out a part of the colony by the offspring (the most unfit). With a small probability we cut out a part of the colony of parents .Well like accidental death .Rambo is not subject to purging of frames
8 This is the end of the era.
9 If during .... (specify ) one and the same individual holds the flag Rambo - the calculations end GA converged, otherwise go to point 3.
....
I wrote my GA with a lot of effort. But it gets stuck in local extremes. Can you give me some general recommendations on how to avoid it?
It's not a contagion, it's a special, "genetic" magic. :)
What function do you use as FF?
I recommend testing, calibrating and debugging your optimisation algorithm on functions specially designed for this purpose.
Please forgive my intrusiveness, but this is a question of principle. I took an example from your article and wrote a script to check it. The answer does not coincide with the example, either the skis are not travelling or the problem is in the padding between the chair and the keyboard. If it is not difficult to poke like a kitten, the second day I can not find the error. Sincerely
Please forgive my intrusiveness, but this is a question of principle. I took an example from your article and wrote a script to check it. The answer does not coincide with the example, either the skis are not travelling or the problem is in the padding between the chair and the keyboard. If it is not difficult to poke like a kitten, the second day I can not find the error. Sincerely
If there is no confidence in the FF itself, in the sense you are not sure that your extrema are such, you can run the function in a tester, tester GA in this sense is quite good, and finds a fairly accurate solution, but only with a small number of parameters (1,2).
In general, it should be understood that the GA finds not an exact, but a robust solution. That is, a solution that is quite good compared to the field of possible solutions.
Please forgive my intrusiveness, but this is a question of principle. I took an example from your article and wrote a script to check it. The answer does not coincide with the example, either the skis are not travelling or the problem is in the padding between the chair and the keyboard. If it is not difficult to poke like a kitten, the second day I can not find the error. Regards.