Algorithm Optimisation Championship. - page 116

New comment
 
Andrey Dik:

Here's the script. I doubt you'll feel any better after what you've seen.

Oh, that's great! Thank God, at least you got something out of it!

So to summarize, - The functionality of the FF library is the same as in the text problem. The same names of exported functions.

Connection interface you have laid out, there is an executable script...

The only thing missing is the FF itself, but I understand how it should be set up and it's not a big deal.

There's a step, there's a range...

Well, in general, it seems to have everything for writing the algorithm.

 
Реter Konow:

Well done! Thank God, at least you got something out of it!

So to summarise, - The functionality of the FF library is the same as in the text problem. The same names of exported functions.

Connection interface you have laid out, there is an executable script...

The only thing missing is the FF itself, but I understand how it should be set up and it's not a big deal.

There's a step, there's a range...

Well, in general, it seems to have everything for writing the algorithm.

You're an odd man.... Many people had the algorithm long before the championship, and no one has any problems. You need an algorithm to participate, and the rest is not important.

And the interfaces, as you can see, are the same as I showed earlier.

 
Andrey Dik:

You are an odd man.... Many had the algorithm long before the championship and no one has a problem. You need an algorithm to participate, and the rest is not important.

And the interfaces, as you can see, are the same as I showed earlier.

Well, those who had the algorithm long before the championship, probably lost it on the way to the championship, that's why they are sitting and keeping silent now...)

You're a weirdo too. Afraid to expose the FF function... What's in it that could help much to beat you?

Here's the FF of text tasks:

double FF (double &param []) export
{ 
  countRuns++; 
  
  int sizeArray = ArraySize (param); 
  if(sizeArray != textLen) 
    return (0.0);
  
  int ffVolue = 0; 
  
  for (int i=0; i< textLen; i++)
  {
    if(GetCode(param [i]) == StringSubstr(Code, i, 1))
      ffVolue++;
  }
    
  return (double(ffVolue));
}

And what's so scary about it that you're afraid to put it out there?

 
Реter Konow:

Well, those who had the algorithm long before the championship probably lost it on the way to the championship, so they sit and keep quiet now...)

You're a weirdo too. Afraid to expose the FF feature.... What's in it that could help much to beat you?

Here's the FF of text tasks:

And what's so scary about it that you're afraid to put it out there?

You do not engage in demagoguery, please.

You and Event have already demonstrated how the text problem was solved...

I have not stated anywhere that I hope or expect to win, I don't care about winning at all. What I want is a comparison. The purpose of the championship is to find the maximum on an unknown FF. My algorithm has no dependence on the FF, just as the MQ optimizer with which the problem was solved has no dependence on the FF. But you have demonstrated a binding to the problem, used knowledge of the key (which you did not cite here), and chose not to answer my leading questions. All this says directly that you are trying to use knowledge of FF, even if not with the aim of winning, but this is contrary to the objectives of the championship, no one should know about FF and should not use this information to solve the problem.

I cannot explain the same thing to you over and over again, we have already written half the forum. Stop being so stupid, I'm really tired of saying and typing the same thing, my keyboard is worn out. Channel your ardor in the right direction. You have already been suspended for such demagoguery but I have kindly agreed to meet you and gave you the green light to take part. Do not lose your last chance. Soon we will have an algorithm ready for you. And Dmitri probably already prepared. Sergiy is ready for sure, but probably on holiday. Better ask them how they have managed (to your surprise) to build their algorithms without knowledge of FF. Ask, ask questions - how can you search for a maximum in an unknown function, read the thread again, see what I said time after time, remember about multidimensional space and other things.... But don't try to influence the championship, please.

 
Andrey Dik:

Don't engage in demagoguery, please.

You and Event have already demonstrated how the text problem was solved...

I haven't claimed anywhere that I hope or expect to win, I don't care about winning at all. What I want is a comparison. The purpose of the championship is to find the maximum on an unknown FF. My algorithm has no dependence on the FF, just as the MQ optimizer with which the problem was solved has no dependence on the FF. But you have demonstrated a binding to the problem, used knowledge of the key (which you did not cite here), and chose not to answer my leading questions. All this says directly that you are trying to use knowledge of FF, even if not with the aim of winning, but this is contrary to the objectives of the championship, no one should know about FF and should not use this information to solve the problem.

I cannot explain the same thing to you over and over again, we have already written half the forum. Stop being so stupid, I'm really tired of saying and typing the same thing, my keyboard is worn out. Channel your ardor in the right direction. You have already been suspended for such demagoguery but I have gone out of the goodness of my heart to meet you and still gave the green light to participate. Do not lose your last chance. Soon we will have an algorithm ready for you. And Dmitri probably already prepared. Sergiy is ready for sure, but probably on holiday. Better ask them how they have managed (to your surprise) to build their algorithms without knowledge of FF. Ask, ask questions - how can you search for a maximum in an unknown function, read the thread again, see what I said time after time, remember about multidimensional space and other things.... But don't try to influence the course of the championship, please.

Don't talk me into remembering about multidimensional space.))

The text challenge has been solved in the most honest way possible.

You constantly confuse (intentionally or not) the notion of FF - program function and the notion of analytical function (key). It is OK to show the software function, but the key (mathematical equation) is not.

I asked to show the FF program function to better present all the nuances of an unfamiliar area of development, and you draw the wrong conclusions about unfairness and an attempt to cheat.

If you remove me from the championship again, I will publicly challenge you to a duel after I finish my algorithm anyway. If you refuse, you're showing cowardice. Better let's have a fair fight and get it over with.

 
Andrey Dik:

Stop being stupid already . You have already been suspended for such demagoguery, but out of the goodness of my heart I went out of my way to meet you and still allowed you to participate. Don't lose your last chance.

Do you have to take that tone? Peter's a big part of making the tournament happen. Last chance... This whole affair has had several chances to sink in. From where it's been occasionally dug up, not even by the contestants. Peter was among them. So we don't know who gave who a chance, because the other "participants" have been giving it all up with persistent regularity.
 
Реter Konow:

Don't be persuaded to remember multidimensional space.))

The text problem was solved in the most honest way.

You constantly confuse (intentionally or not) the concept of FF - program function and the concept of analytical function (key). It is OK to show the software function, but the key (mathematical equation) is not.

I asked to show the FF program function to better present all the nuances of an unfamiliar area of development, and you draw the wrong conclusions about unfairness and an attempt to cheat.

If you remove me from the championship again, I will publicly challenge you to a duel after I finish my algorithm anyway. If you refuse, you're showing cowardice. Better let's have a fair contest and end this thread.

If you want to compete, finish your algorithm already. You can't compete with something that doesn't exist. When it's ready, we'll have something to talk about. If you want to know how to get the most out of an unknown feature, ask, maybe someone will give you an answer.

And stop trying to measure what you don't have.

Go to the Literature folder. There is a whole section on optimization algorithms. Read it, get ready, write an algorithm, such things can't be done on the spot, you need deep knowledge and more than one day of hard work.

 
Andrey Dik:

1000 accesses. Initialisation by random values in the range.

It's not clear to me why the first call has values of 0.5? Why not 0.01 or 57.19 or 100.0? Very interesting.

D The question of the first reference to the FF is an important one, but not the main one. It depends on the range of search and number of parameters. In my algorithm, it can be random with a given distribution.

The main thing for me now is something else.

I compiled an algorithm for searching for a minimum and took the function F(x1,x2,x3)=exp(x1+x2+x3)/(x1*x2*x2*x3*x3) to check its performance. Let's compare the results.

My resultNumber of times the function is called - 51Min=3.76210

Your result Number of times the function is called1000Min=3.7354.

The question is which of us won? I, which calculated the minimum with given error with smaller reference to the FF. Or you, who calculated more exact minimum but for much more calls to the FF?

It is fundamentally important for me for adjustment of the algorithm. My opinion - these parameters are equally important. I may think about coefficients of the algorithm's overall estimation.

I ask refereeAnatoli Kazharski to comment.

 
Yuri Evseenkov:

The question is which of us wins?

Is the true minimum known? The answer determines how the results table I gave here is calculated.

The accuracy criterion is 3 times weightier than the time criterion.

Instead of time, we should use the number of runs (time spent calculating the FF is proportional to the number of runs).

 
Yuri Evseenkov:

D The question of the first reference to the FF is an important one but not the main one. It depends on search range and number of parameters. In my algorithm it can be random with a given distribution.

This question is very important. Initial parameter initialization directly affects convergence of the algorithm. And if the FF is unknown, then only initialization by random numbers is suitable.

1...109110111112113114115116117118119120121122123...133
New comment
Reason: