Algorithm Optimisation Championship. - page 14
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
The space can only be three-dimensional. My imagination refuses to imagine otherwise. As I understand it, a single FF is not suitable for a championship because its surface is too simple. To complicate the surface do you want to use multiple FFs? By superimposing the curves they create on top of each other would you create a surface complex enough to test the algorithms for universality?
All right. Let's have a three-dimensional search space, if you insist.
The task will have between 100 and 500 optimisable parameters, and this is the main thing.
IMHO. The discussion has gone off into the distance...
How about starting the first round of the championship with a simple task that everyone can understand?
For example, a simple example:
Find the roots of the equation: 34a+43b+16c+30d+23e=4492;
All algorithms can be used: brute force, evolutionary, pre-revolutionary...
Participants solve an equation given by the organizer. The fastest and most accurate answer is the winner.
The discussion is free to go into the distance of any distance, and the championship awaits uson 11 July 2016.
In the example you presented there are 5 variables, or according to you 5 roots. Can you solve levels with 100 roots? And with 500? - If yes, welcome to the championship!
I would like to participate, but for me it is acceptable only so far that the organizers will lay out the data and ask to solve in any way and in any language.
You don't need to find all the bumps, one more or less high is enough. It's an optimization task, not a matan task.
Yes. You don't have to. We need to find as much FF championship value as possible.
What is matan?
The question is - how far in my understanding (following the accepted analogies) do I deviate from the subject of the championship?
Why ask questions like that? - whether you do or don't.
The task is to find the maximum of an unknown function with parameters between 100 and 500. That's all you need to know to enter the championship. It doesn't matter how many dimensions the search space actually is, it's the answer that counts. Find it!
The discussion is free to drift off into the far reaches of any distance, and the championship awaits uson 11 July 2016.
There are 5 variables in the example you presented, or according to you 5 roots. Can you solve levels with 100 roots? And with 500? - If so, welcome to the championship!
Easy. By full brute force. But this is the longest option. There can be as many roots as you like. You can do that. The point is to see who can find the roots faster and more accurately. And you don't need to send algorithms.
Easy? Great!
How do you check "faster" and "more accurately" if the algorithms are in the hands of the participants? How do you check that a participant has found a solution in fewer steps than a complete brute force?
Obviously, optimization has something to do with search. That is, instead of a complete search, the search applies a strategy to find the desired value faster. It is quite clear that the surface node search can be optimised. But what is meant by optimization of parameters? Parameters of what?
Function parameters.
Ok, let me put it this way, your algorithm should be able to pass values, parameters as an array to a library with FF and get back a response this way:
double FF(double &array []);
This is very easy to understand, isn't it? Array[] has dimensions from 100 to 500 (to be solved later). The task of the participants is to find such values in the array, at which the FF response will be maximal.