Algorithm Optimisation Championship. - page 67
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
In general, the idea is that everyone can post their own FF as ex5 here.
Anyone will be able to check the results. Just download the published FF and run search for its maximal value in your script. The result can be posted here, but must be accompanied by the algorithm also as ex5, so that anyone can check his finding.
Then you don't need to upload the library of the algorithm. Everyone calculates the maximum and post the result with the found parameters (which are easily checked, just paste the parameters into FF and verify the result). Problems of preserving intellectual property rights are eliminated.
But there is a problem, for how many runs was the result obtained? - it will not be known, which is almost the most important thing.
Then it is not necessary to post the library of the algorithm at all. Everyone will calculate the maximum and post the result with the found parameters (which are easy to check, just paste the parameters into FF and check the result). Problems of preserving intellectual property rights are eliminated.
Mr Zielinski simply has his finger on the pulse of our championship. When he sees that the championship is "suffocating", he rushes to the rescue with a "defibrillator". Thanks to him for that. :)
But seriously, everything is almost ready for the championship.
There are participants, there is an understanding of the task, there are methods of connection, the FF is very easy to do...
All you need is an agreement to make decisions jointly. There is nothing else.
You, Andrei, just leave your supremacy and become a participant like everyone else.
Then the championship will come true.
Mr Zielinski simply has his finger on the pulse of our championship. When he sees that the championship is "suffocating", he rushes to the rescue with a "defibrillator". Thanks to him for that. :)
But seriously, everything is almost ready for the championship.
There are participants, there is an understanding of the task, there are methods of connection, the FF is very easy to do...
All you need is an agreement to make decisions jointly. There is nothing else.
You, Andrei, just leave your supremacy and become a participant like everyone else.
Then the championship will come true.
Carry it out. I gave you a push, showed you what and how you can calculate and check, explained the problems and ways to overcome them when estimating the quality of algorithms. I showed you sample codes and an example of a working algorithm. He provided me with exhaustive information in the form of many books on the subject.
Further - my participation is not required. Everyone be active and move the championship forward. Compete. I have been reproached many times, including by you, that my algorithm is "too strong", so I will not embarrass you with the presence of my algorithm.
Carry out. I gave a push, showed what and how you can calculate and check, explained the problems and how to overcome them when assessing the quality of algorithms. I showed examples of codes and an example of a working algorithm. He provided me with exhaustive information in the form of many books on the subject.
Further - my participation is not required. Everyone be active and move the championship forward. Compete. I have been reproached many times, including by you, that my algorithm is "too strong", so I will not embarrass you with the presence of my algorithm.
You, Andrei, once said that my presence at the championship is important in principle. Now I'm telling you the same thing. About your algorithm being too strong, I have reasons to think otherwise. I can justify it.
Justify it, if you're not too lazy.
Justify, if you're not too lazy.
1. No one has proven that the "evolutionary" approach of finding optimal parameter values is the most efficient.
2. Even if this approach is the most efficient, it is not a fact that you have reached perfection in it.
3. You reacted painfully to suggestions to change the range and pitch. It was obvious that your algorithm was already set up and changing such important parameters could disorient it.
4. Finding optimum values, is not always a search for the maximum. The optimal value can be "on the slope". You have always only ever talked about the maximum values of the FF. (Again, perhaps because that's how your algorithm is set up).
5. You did not give the right analogies to understand the problem. Even in the disappeared post that I saw, you were talking about multidimensional space. However, it is quite clear that this problem has nothing to do with this theory. Perhaps you yourself do not clearly understand the essence of optimization.
6. There are very big doubts whether you understand FF correctly, since in no way is there a need for additional coordinate axes. There is only one curve. It is constructed by a formula in which the parameters can be as many as you like.
The point is that we know all these parameters. At any moment of time, we substitute their values into the formula and get the coordinates of a certain point of the curve line in the graph.
Our task is to quickly find those values that give the optimal point on the graph.
We don't need more than two measurements to solve the problem.
These are the reasons for my doubts about the efficiency of your algorithm.
1. No one has proven that the 'evolutionary' approach of finding optimal parameter values is the most efficient.
2. Even if this approach is the most efficient, it is not a fact that you have reached perfection in it.
3. You reacted painfully to suggestions to change the range and pitch. It was obvious that your algorithm was already set up, and changing such important parameters might disorient it.
4. Finding optimum values, is not always a search for the maximum. The optimal value can be "on the slope". You have always only ever talked about the maximum values of the FF. (Again, perhaps because that's how your algorithm is set up).
5. You have not given the right analogies to understand the problem. Even in the disappeared post that I saw, you talked about multidimensional space. However, it is quite clear that this problem has nothing to do with this theory. Perhaps you yourself do not clearly understand the essence of optimization.
6. There are very big doubts whether you understand FF correctly, since in no way is there a need for additional coordinate axes. There is only one curve. It is constructed by a formula in which the parameters can be as many as you like.
7. The point is that we know all these parameters. At any moment of time, we substitute their values into the formula and get the coordinates of some point of the curve line in the graph.
Our task is to quickly find those values that give the optimal point on the graph.
We don't need more than two measurements to solve the problem.
These are the reasons for my doubts about the efficiency of your algorithm.
No, it's not justification, it's bullshit. And it is total.
1. Not only GA and its derivatives are called evolutionary, but all that can optimize functions whose formula is unknown. Therefore my algorithm is as evolutionary as any other algorithm presented at the championship.
2. It is not a fact that my algorithm is the most perfect, and the opposite is not a fact either. So it's wrong to say "I doubt your algorithm is too strong", the correct thing to say is "the search ability of your algorithm is not known".
3. what a load of crap. There is no painfulness. You can choose any range and pitch and I justified in the post that mysteriously disappeared, to the content of which you are now appealing, that it doesn't matter range and pitch, they can be any, all that matters is that we can't set 0th pitch due to restrictions on intellectual property protection.
4. The maximum is how the FF is represented, it has been said before. It is always possible to represent a function in such a way that any point becomes "maximal". Finding the maximum is not a limitation of the algorithm, it is a feature of the FF representation.
5. What do you mean "didn't give the right analogies"? Why would I give them at all? Analogy is such a thing that it is necessary for someone who doesn't understand the essence of the phenomenon, not for someone who does. Ok. I'll ask you a puzzling question, your answer to which will give a full indication of your level of training: "Which algorithms will get the advantage and why for FF(f1(x1,y1)+...+ f250(x250,y250)?"
6. It doesn't matter how we imagine FF. What matters is what kind of representation will help me find a solution. I have described why a different representation will help me, and you may find a different representation, e.g. in the form of elephants, and I will not claim that your representation is wrong.
7. I've already grimaced today. Tomorrow I'll drop the FF here in the form of a library. See if you can find a solution. Everyone is welcome to try it. Let's see what the words "parameters are known" are worth.
No, it's not justification, it's bullshit. And it is a complete one.
1. Not only GA and its derivatives are called evolutionary, but all that can optimize functions whose formula is unknown. Therefore my algorithm is as evolutionary as any other algorithm presented at the championship.
2. It is not a fact that my algorithm is the most perfect, and the opposite is not a fact either. So it's wrong to say "I doubt your algorithm is too strong", the correct thing to say is "the search capabilities of your algorithm are unknown".
3. what a load of crap. There is no painfulness. You can choose any range and pitch and I justified in the post that mysteriously disappeared, to the content of which you are now appealing, that it doesn't matter range and pitch, they can be any, all that matters is that we can't set 0th pitch due to restrictions on intellectual property protection.
4. The maximum is how the FF is represented, it has been said before. It is always possible to represent a function in such a way that any point becomes "maximal". Finding the maximum is not a limitation of the algorithm, it is a feature of the FF representation.
5. What do you mean "didn't give the right analogies"? Why would I give them at all? Analogy is such a thing that it is necessary for someone who does not understand the essence of the phenomenon, not for someone who does. Ok. I'll ask you a puzzling question, your answer to which will give a full indication of your level of training: "Which algorithms will get the advantage and why for FF(f1(x1,y1)+...+ f250(x250,y250)?"
6. It doesn't matter how we imagine FF. What matters is what kind of representation will help me find a solution. I have described why a different representation will help me, and you may be helped by a different representation, e.g. in the form of elephants, and in doing so I will not argue that your representation is wrong.
7. I've already grimaced today. Tomorrow I'll drop the FF here in the form of a library. See if you can find a solution. Everyone is welcome to try it. Let's see what the words "parameters are known" are worth.
1. So an evolutionary algorithm is also called a complete enumeration of all values?
2. That's exactly the point I was making in this paragraph. You repeated the meaning of what I said.
3. About painfulness - my subjective impression, but you fundamentally refused to consider changing the range and pitch.
4. True, that is how the FF is presented. But the optimum value may not be the maximum value. You never mentioned it. But you could have.
5. No expert in his field would give analogies for understanding which could be misleading. That is obvious. Your analogies were from the realm of science fiction.
6. At no point have you clearly articulated what FF is. Your explanations have contributed to such a wide range of perceptions in such a wide field of imaginative play that one could not help but question your own understanding of what you are talking about.
All of the above explains my doubts about your competence on the subject, but it doesn't mean that you are really incompetent. Perhaps you are just that kind of expert.
Regarding the fact that you don't have to explain anything to others and are doing them a favor by explaining it, I agree with you. You do.