How do you work with neural networks? - page 2

New comment
 
luka:
What about training? Training can be done at the start

You can do it in the end, that's not the point. Where and how? Having the nets themselves in an expert is not a problem, but training?

 
This is how, in my humble opinion, a capable network should be constructed:


[...]
In addition to the circuit in charge of such movements, each black
crystal contained another circuit, or rather a fragment of it.
a fragment, because it seemed to be part of some
larger structure. This supreme whole, probably arising
only when a huge number of elements come together, was
the true motor that set the cloud in motion. Here, however,
the scientists' knowledge was cut off. They didn't know the
the growth potential of these super-systems, and it was very unclear
was the question of their "intelligence". Cronotos conceded that
the more elementary units joined together, the more difficult
the more difficult the problem they had to solve. This sounded pretty convincing, but
neither cyberneticists nor information theorists knew
anything consistent with such a construct, i.e. an "arbitrarily
a "sprawling brain" that fits its size to
the size of its intentions.
[...]
Stanislav Lem. "Invincible".


So far, I have had only modest success in constructing the "crystal".

 
sayfuji:
Did you achieve this result using Neuro Shell Day Tarder? If it is not a secret which version. I think there is no possibility to import dll. Then how does it communicate with MT4, or rather the "code aspect"?

This is exactly what has been achieved in Neuroshel. The version does not matter, you can work on any version - as long as you know how. The differences in versions are minimal. And this does not mean that the higher the version the grander. But 5.5 beta 4 is the last. They have no connection with MT4, they just rewrote the network from the Neuroshel (C+) addon to MQL. I tried to optimize it in MT but the optimizer shows so many optimization variants that it would burrow too much. I cannot find a correct variant, at least, not at me.

 
What is the addon and where to download it? I'm interested in how to rewrite my strategy from NSDT to mql4.
 
So what's the addon? And what do you use to look at the code that's in C+?
 
sayfuji:
So what is the addon? And what do you use to look at the code, which is in C+

The addon is listed in that branch. The code was given to me by the developers.

 

LeoV 28.04.2008 14:03
Figar0 писал (а):

З.Ы. На всякий случай поинтересуюсь, это разработка в лоб на MQL или, зная Ваше интерес к NSDT, с его использованием?

Конечно с использованием NSDT. Вся оптимизация в NSDT. Хотя в МТ4 лучше получается подобрать лоссы и профит.....

Reread the thread but couldn't find the addon in question. It only said adaptive neural network. If you're comfortable, please send information on how to view the code from the strategy in NSDT to sayfuji dog ya.ru or just here


 
sayfuji:

Reread the thread but couldn't find the addon in question. It only said adaptive neural network. If it's convenient for you, send info on how to view code from the strategy in NSDT to sayfuji dog ya.ru or just here

So it is an addon. I don't even know how to see the code in the strategy.....

 

But did you somehow create an EA that takes results from the NSDT, or did the developer just give you the source code for the adaptive grid?

 
sayfuji:

But did you somehow create an EA that takes results from NSDT, or did the developer just give you the source code for the adaptive grid?

Taking results from the NSDT is one thing. There are different addons for that. And sources is another. In this case, I used the sources.

12345678
New comment