Algorithm Optimisation Championship. - page 109

 
Yuri Evseenkov:

Minimum Fmin=3.76210

x1=1.1; x2=2.1; x3=3.1;

Can someone check the minimum?

The minimum is 3.73545179.

you have to count starting from 0.01 or else it will divide by 0.

only why is the first referencex1=x2=x3=0.5;?

 
ILNUR777:
It's not about whether I need it or not. The matter is that you write that there are a lot of examples and it turns out that there are none. Of course, I didn't write about dumb applicability of optimization to anything. Shovel is also a tool, but if you do not know where the money, then the money will not dig, but for it can dig anything, whether it means that people can not dig a trench at speed - no. Who has fun?

Read my post about the scientist and the microscope. You must have missed it. I told you that optimisation does not make a profit on its own, but that you cannot make a profit without it. Just as an engine does not drive itself, it can only drive as part of a car.

It feels like you are talking for the sake of talking. If you need optimization - use it, if you do not need it - do not use it. What is the problem? There is a championship of optimization algorithms here, a competition for accuracy and minimum reference to the FF. As analogy with cars, it is not the championship of rally or formula 1, it is the championship of engines for efficiency, the winner is the one who will develop more power and will burn less petrol at that.

 
Bolkonsky, FAS!
 
Yousufkhodja Sultonov:
What do you mean by"optimization criteria"? I thought there is a single optimization criterion - a set of optimization parameters (the less the number - the better, and ideally, one - period) that provides the maximum value of recovery factor FS = ratio of net profit to the maximum drawdown.

The optimisation criterion has always been the indicator that the optimisation is seeking to improve. If optimised by the profit criterion, the aim is to increase it, if optimised by the loss criterion, the aim is to reduce it.

The number of parameters is a characteristic of the system. Of course, the system can also be optimized by the criterion of reducing the number of parameters, but in such a way that it would then allow optimization by the criteria (indicators), for the achievement of which the system is designed (profit, loss).

 

What's up with the FF?

I have a feeling I won't be able to continue my work properly until this thread is over...)

We need to start the competition soon.

 
Реter Konow:

What about FF?

I have a feeling I won't be able to continue my work properly until this thread is over...)

We need to start the competition soon.

I gave you two FF formulas, train on them. And how will you train/compete if you don't have an algorithm yet?

And Yuri already showed the problem, solve it.

 
What's wrong with the mobile version of the site? When I reply to a post, the quote code appears instead of the text of the quote. To Andrew: How many times did you access the function to get this minimum? My first access to the function was as in the description of the method.
 
Dmitry Fedoseev:
Bolkonsky, FAS!
 
Andrey Dik:

I gave you two FF formulas, train on them. And how will you train/compete if you don't have the algorithm yet?

And Yuri has already shown a problem, solve it.

I want to create an algorithm.

To do this, I need an FF library that includes the entire set of exportable functions, including the FF and the connection interface. The connection interface you have already laid out.

FF, as you yourself know, is not an analytic function, but a function that includes an analytic function, in which an array with parameter values is substituted.

Without the FF and the entire library, I cannot create an algorithm to "train" and compete.

Given that you yourself know it all, and I know that you know it all, your advice to "train" sounds somewhat ridiculous.

 
Реter Konow:

I want to create an algorithm.

To do this, I need an FF library, which includes the entire set of exportable functions, including the FF and the connection interface. You have already laid out the connection interface.

The FF, as you yourself know, is not an analytic function, but a function that includes an analytic function in which an array with parameter values is substituted.

Without the FF and the entire library, I cannot create an algorithm to "train" and compete.

Given that you yourself know it all, and I know that you know it all, your advice to "train" sounds a bit ridiculous.

So, what difference does it make which function? F(x1,x2,x3, ...xn). It only matters how many variables, and that to simplify things, because the program will need different software for an arbitrary number. If analytical, that alone is enough.

Here's the only thing, when you combine it all - this F(x...), somehow you have to make sure beforehand that it only has one max). If I'm not mistaken, I don't think anyone has asked that question. And then you have to look for several and choose the coolest one), and that's another task again.