I will write a bespoke expert - page 6

 
Vinin писал (а) >>

Is this the same or not?

http://library.graphicon.ru/pubbin/view_prop.pl?prop_id=19

No.

I "took" (and played with) NS2 when I was trying to figure out if I could "get" something from the inputs "out of the box" by picking up the architecture. :)

If "positioning", then (from the help):

NeuroShell 2 incorporates a very powerful architecture called the Argument Group Accounting Method (MGUA) or polynomial networks. In fact, an MGUA network is not like regular direct-coupled networks, and originally this architecture was not usually represented as a network. An MSUA network contains polynomial expressions in its links, and uses in some sense a mechanism similar to genetic algorithms to decide how many layers to build. The result of the training is the ability to represent the output as a polynomial function of all or part of the inputs.

MGUA was invented by Prof. A.G.Ivakhnenko of the Institute of Cybernetics, Academy of Sciences of Ukraine, Kiev, Ukraine. A more detailed description in English of the design and applications of the MSUA can be found in a book edited by Farlow (see References). (For reading in Russian, besides numerous articles by A.G.Ivakhnenko with co-authors, one can recommend his books, the outputs of which are also given in the references. - (Translator's note).

'

The basic idea behind MSUA is that the algorithm tries to construct a function (called a polynomial model) which behaves in such a way that the predicted output value is as close as possible to its real value. For many users it is very convenient to have such a model capable of making predictions using familiar and easy to understand polynomial formulas. In NeuroShell 2 the MSUA is formulated in terms of a neural network architecture called a polynomial network. However, the resulting model is a standard polynomial function.

The most traditional approach for solving the problem of building such models is to use regression analysis. The first step is to decide on the type of polynomial to be constructed by regression.

'

Literature

...

Farlow, S. J. (ed.), Self-organizing methods in modeling: GMDH Type Algorithms. Statistics: Textbooks and Monographs, 54, 1984.

 

At one time I too was actively studying MSUA, I even wrote an EA for MT.

But in the end the Expert Advisor turned out to be a failure... Maybe I messed up something with inputs and outputs, or maybe I was forming data that contradicted each other, and that's why MHUA was unable to find a decent polynomial... In general, this MSUA only frustrated me.

On the other hand, maybe my brain wasn't praised...?

 
Vinin писал (а) >>

You can go into more detail here.

What kind of detail? Describe the method?

You can read the description at: http://www.codenet.ru/progr/alg/ai/htm/gl3_7.php

In general, there is a lot of information on this subject, but I wonder if it is worth trying to implement or not. I would like to know if anyone has tried it and how successful it was.

 
worked with genetic algorithms. MSUA is the same as GA in principle. It is also an evolutionary search.
 
3 years ago
 
njel писал (а) >>
worked with genetic algorithms. MSUA is the same as GA in principle. It is also an evolutionary search.

I don't agree with that. GA is much simpler to implement. GA doesn't have much in the way of mathematical formulas or complex transformations.
But MGUA... A lot of things are described in a vague or too general way, so sometimes it's hard to imagine how it can be programmed.

Does anyone else have any experience in this area?

 
Lemyx писал (а) >>

I don't agree with that. GA is much easier to implement. GA doesn't have any mathematical formulas or complicated transformations.
But MGUA... Many things are described too vaguely or too generalized, so sometimes it's hard to imagine how it can be programmed.

Does anyone else have some experience in this area?

I have some work, but it's in C Builder, left over from uni, I can share if you need it:)

 
slayer писал (а) >>

а кто мне напишет такого советника, который будет файлы в формате *.ex4 декомпелировать в формат *.mt4 ????

СЛАБО???? (((

slayer
писал (а)
>>

Sorry *.mq4

http://www.procapital.ru/showthread.php?t=8065

 

Do you use it yourself? I don't really believe that such a golden thing costs 5 c.u.:)

 
Lyfesh писал (а) >>

Do you use it yourself? I don't really believe that such a golden thing costs 5 c.u.:)

Part of the advertising -- watch the number of posts, and you also reproduce the link for the search engines... And on the subject, a couple of three of them will get hooked.