To follow up - page 32
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 the quest for clustering, we work with fixed points in time, whether it's a perfect input-output, a benchmark trading system or something else.
Music of the Spheres. It is good to hear that the method of fixing these moments is not important. Especially since the ideal inputs-outputs are given by one of the benchmark systems. I may add that the "total balance substitution" causes the balance surface constructed in the optimiser on the whole FP to lose its physical meaning. The points on this surface are incomparable with each other, as they differ not only in parameter values, but also in input-output sets. As a result, the value of a particular set of parameters cannot be determined - at a different segment of history, a directly opposite result can be obtained for it and the whole surface will have a completely different appearance.
In the case when the input-output system is fixed by some algorithm the situation becomes quite different. The balance turns out to depend only on a set of parameters, and this gives hope that the clustering will not change its statistics when moving to another piece of history. To do this, of course, it has to be built on sufficiently large data sets. So you always need a pair of input-output algorithms and an FP parmetrization to build a TC.
By the way, if the input-output algorithm has its own parameters, the task becomes more difficult. However, even in this case, the proper use of FP allows us to get results.
Re-read the entire branch. Well, I don't remember anything like that lately. Just an awesome thread. Ideological disputes between morganists and veismanists are almost irrelevant, these bourgeois are all the same.
Музыка сфер. Приятно слышать, что метод фиксации этих моментов неважен. Тем более, что идеальные входы-выходы дает одна из эталонных систем. Могу добавить, что "подмена итогового баланса" приводит к тому, что поверность баланса, построенная в оптимизаторе на всем ФП, теряет физический смысл. Точки этой поверхности несравнимы между собой, поскольку они отличаются не только значениями парамтеров, но и множествами входов-выходов.
No, it wasn't. Balance swapping is just to slip the optimizer different results for the same set of inputs-outputs. Simply compare two-dimensional diagram of optimization results with my diagram and it will become clear to you what to replace for one to turn into another.
However, once again, I don't see any prospects for such an approach.
About the music of spheres - I meant only irrelevance to the question posed :)
P.S. Alexey, why don't you consider a possibility to output 2D diagram just in indicator window? People have learned how to draw almost everything in MT (all two-dimensional at least :) ).
No, it isn't. The purpose of balance swapping is to give the optimiser different results for the same set of inputs-outputs. Just compare two-dimensional diagram of optimization results with my diagram and you will see what needs to be replaced to turn one into the other.
I didn't doubt that you'll be against it. :-)
Maybe, what I wrote is not clear to you, because your input-output system (pardon - trades) is fixed initially, and therefore what I said does not concern your version at all ?
Well, it doesn't matter. Be that as it may, I have no intention to get involved in another argument with you.
P.S. Alexey, why don't you consider displaying a two-dimensional diagram simply in the indicator window? People have learned how to draw practically everything in MT (all two-dimensional at least :) ).
As a valid and authorized representative of the people I confirm. Here, for example, is the surface depicted in colour over a two-dimensional FP.
Long live MT ! Or MQL ? А ... MetaQuotes !
Не сомневался, что ты будешь против. :-)
Может быть то, что я написал, неочевидно для тебя потому, что у тебя система входов-выходов (пардон - сделок) зафиксирована изначально и поэтому сказанное мной к твоему варианту вообще не относится ?
Впрочем, это неважно. Как бы там ни было, у меня нет намерения втягиваться в очередной спор с тобой.
Как действительный и полномочный представитель народа подтверждаю. Вот, например, поверхность изображенная цветом над двумерным ФП.
Да здравствует МТ ! Или MQL ? А ... MetaQuotes !
Pardon me. :[ How do you do that?
There is no need for optimisation here. A single test run is enough, and along with each transaction you must output a unique vector of context parameters. And to cluster in Excel or somewhere else. But to draw it would be interesting and even beautiful.
Anyway, the point is that it would be nice if in MT6 (or maybe even in MT5) the optimizer had been given structure flexibility enough to be able to mold a tool that could handle different and multi-step data processing.
P.S. Two-dimensional manifold is not very difficult to draw. But what if it turns out to be multidimensional (context parameters are more than two)? This is where Kohonen comes in handy, but I still need to get into it...
P.S. Двумерное многообразие нарисовать не очень сложно. А если там многомерное получится (параметров контекста больше двух)? Ну тут прям Кохонен так и просится, но мне еще въехать в него надо...
The networks first of this thread are begging here.
P.S. A two-dimensional manifold is not very difficult to draw. But what if it is multidimensional (context parameters are more than two)? This is where Kohonen comes in handy, but I still need to get into it...
Kohonen is not so complicated. Besides it may be done in Excel.
Короче, дело идет к тому, что неплохо бы в MT6 (а, может быть, и уже в MT5) наделить оптимизатор достаточной гибкостью структуры, чтобы уметь лепить из него инструмент, способный на разные и многоступенчатые обработки данных.
It is sufficient to add the option of specifying an arbitrary fitness function in the optimiser. Then such tasks can be clicked like nuts.
P.S. A two-dimensional manifold is not very difficult to draw. But what if it's multidimensional (context parameters are more than two)? Well, Kohonen is really asking for it, but I still need to understand him...
Nah, Kohonen is not very good here. IMHO, networks based on radial basis functions are better suited here.
Oftop, sorry. :[ How do you do that?
Is it a question or an exclamation ? :-)
Неее, Кохонен тут не очень. ИМХО, тут сети на основе радиальных базисных функций лучше подойдут.
Это вопрос или восклицание ? :-)
Probably both. :)