Algorithm Optimisation Championship. - page 11
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
Can someone really come forward with an ALGLIB library?
Hello all.
I have formed a few questions on the technical details.
1. in what form will the FF be implemented? - EA, script, library?
2. I assume that the FF will be implemented as an EA (i.e. it will be inside the EA), and from there it will call the member library with the optimization algorithm.
If so, is the number of calls to the algorithm limited?
3. If point 2 is correct, then the FF will call the algorithm, which in turn will call the FF. Suppose the number of calls to the participant's algorithm by the FF is limited.
Is the number of calls back to the FF from the algorithm's side limited (otherwise there would be a closed, infinite loop of calls between the FF and the participant's algorithm)?
4. Who will write the function creating the FF value area? Obviously, it must not be done by the participant. How can we be sure of its neutrality?
5) Where will the peak values of the FF be stored (the maximums we are looking for)? In the array of the FF itself? How can we then check the accuracy of the coincidence between the FF maxima and the values calculated by the contestant's algorithm?
6. Where will the results of the participant's algorithm be recorded and stored?
Hello all.
I have formed a few questions on the technical details.
1. in what form will the FF be implemented? - EA, script, library?
2. I assume that the FF will be implemented as an EA (i.e. it will be inside the EA), and from there it will call the member library with the optimization algorithm.
If so, is the number of calls to the algorithm limited?
3. If point 2 is correct, then the FF will call the algorithm, which in turn will call the FF. Suppose the number of calls to the participant's algorithm by the FF is limited.
Is the number of calls back to the FF from the algorithm's side limited (otherwise there would be a closed, infinite loop of calls between the FF and the participant's algorithm)?
4. Who will write the function creating the FF value area? Obviously, it must not be done by the participant. How can we be sure of its neutrality?
5) Where will the peak values of the FF be stored (the maximums we are looking for)? In the array of the FF itself? How can we then check the accuracy of the coincidence between the FF maxima and the values calculated by the contestant's algorithm?
6. Where will the results of the participant's algorithm be recorded and stored?
Hi.
1. library *.ex5.
2. See point 1.
3. Above in the discussion we discussed that there will be two types of FF library call (at participant's choice) - from executing script or from participant's library with AO.
4. previously discussed. Already after the participants' libraries for the championship will be a discussion and the formation of the FF. So that the FF was controlled by one of the participants will not work, do not worry about it.
5. They will not be stored anywhere. The FF is an analytical function in the form of formulas, the values of the global max will be known only after the "discussion and formation of the FF".
6. All intermediate results or one final result (the fact of one or the other will be taken into account in the final evaluation of the algorithm) the participant's algorithm shall save in *.csv file. Based on these data from the file, the algorithm will be evaluated according to the criterion of "accuracy". Naturally, the algorithms will not know the real global max, see item 5.
Examples of codes and explanations of the connection will be provided on June 20, said earlier. By that time you should have a ready working algorithm, which should be prepared according to the voiced interfaces to the test program from June 20to July 11.
Hi.
1. library *.ex5.
2. See point 1.
3. Above in the discussion, we discussed that there will be two types of FF library call (participant's choice) - from the executing script or from the participant's library with AO.
4. previously discussed. Already after the participants' libraries for the championship will be a discussion and the formation of the FF. So that the FF was controlled by one of the participants will not work, do not worry about it.
5. They will not be stored anywhere. The FF is an analytical function in the form of formulas, the values of the global max will only be known after the "discussion and formation of the FF".
6. All intermediate results or one final result (the fact of one or the other will be taken into account in the final evaluation of the algorithm) the participant's algorithm shall save in *.csv file. Based on these data from the file, the algorithm will be evaluated according to the criterion of "accuracy". Naturally, the algorithms will not know the real global max, see item 5.
Examples of codes and explanations of the connection will be provided on June 20, said earlier. By that time, you should have a ready working algorithm, which should be prepared according to the voiced interfaces to the test program from June 20to July 11.
3. If the FF call types are already stipulated, now I have to decide, should I write my algorithm in a script or in a library? (If I write my algorithm in a library, and the FF itself is also a library, how will the two libraries call each other? I don't get it.)
4. If the FF formation will take place after the participants have finished writing their algorithms, then it means that the algorithm should not depend on a particular FF and be universal?
The participant's algorithm should be ready before the connection examples are provided?
Then the discussion and writing of the FF should be after making a connection to it.
And immediately after establishing a connection to the FF library, the participant must lose access to their algorithm. After that, the FF can be composed. (otherwise, knowing the FF, the participant can change his algorithm).
Have I got it right?
This is the perception that has emerged:
My algorithm transmits some values to the FF, and in response receives some values from the FF.
Then my algorithm analyses values received from the FF and searches for "global" maxima of the FF itself, relying on the logic of values received from it.
Thus, my algorithm approximately calculates 100 or 500 maxima of FF, and saves results in a file.
The more accurate the coincidence of my AO's calculation results with the FF's peak values, and the fewer references to the FF he makes, the higher his place in the championship results will be.
????
3. if the FF call types are already stipulated, now I have to decide, should I write my algorithm in a script or in a library? (If I write my algorithm in a library, and the FF itself is also a library, how will the two libraries call each other? I don't get it.)
4. If the FF formation will take place after participants finish writing their algorithms, it means that the algorithm should not depend on a particular FF and be universal?
The participant's algorithm should be ready before the connection examples are provided?
Then the discussion and writing of the FF should be after making a connection to it.
And immediately after establishing a connection to the FF library, the participant must lose access to their algorithm. After that, the FF can be composed. (Otherwise, knowing the FF, the participant can change his algorithm).
Have I got it right?
3. Yes, call the library from the library if you choose to do the second type of connection.
4. Quite right. The idea, the requirement of universality, independence from the FF I've been trying to get across for a long time. That's what should be, hmm.... amazing ability of the algorithm to work with any FF.
That's the notion that's formed:
5. My algorithm passes some values to the FF and receives some values from the FF in response.
6. Then, my algorithm analyses the values it receives from the FF and looks for the "global" maxima of the FF itself, based on the logic of the values it receives from it.
7. Thus, my algorithm approximately calculates 100 or 500 maxima of the FF and saves the results in a file.
8. The more accurate is the coincidence of results of calculation of my AO with peaks of FF, and the less appeals to the FF he makes, the higher will be his place in the championship results.
????
5. Yes.
6. Yes.
7. Yes.
8. Yes.
I get the impression that you don't have an algorithm yet. Hurry up, there is not much time left. First, write an algorithm, or take a ready one, and then, according to the requirements of the Championship, set up a connection to the test script after June 20.
Will the organizers have a benchmark, something to compare the quality of the algorithm with ? Otherwise, all attempts will be worse than the built-in MT genetic algorithm.