Neural networks, how to master them, where to start? - page 2

New comment
 
nord >> :
I'm interested in neural networks too, I've read and understood them... I tried to tinker with the probability network in codebase libraries, but nothing came out... you have to start from the simplest one, it's a bit complicated.... You know neural networks cannot tell for sure if the price will go up or down? ....mdah, you should be able to use them

So you should try to attach a simple neural network to the most primitive EA right here on the forum... I offered the algorithm with fractal. I can show you the code of advisor with detailed comments in the code.

 

neural networks are just like a computer - you use = know))) thank you for a happy childhood
it's useless and painful to use books.
all files, including perseptron, are extracted ready-made from some generator.

so you need either neuroshell, or Statistics 8, or matlab/matcad with the NN package.

 
Korey >> :

neural networks are just like a computer - you use = know))) thank you for a happy childhood
it's useless and painful to use books.
all the files, including the perseptron, are extracted off the shelf from some generator.

so you need either neuroshell, or Statistics 8, or matlab/matcad with the NN package.

Korey, Well, when you write messages in this thread, you should take into account that it is mostly read by very green users.

 
Korey писал(а) >>
It's useless and painful from books.

It's a lot of work, but you need a basis and understanding, even if without details, otherwise it's a poking and prodding method. And you can poke all the way to the carrot's wedding, it's too vast an area...

 

to Andrey4-min

Well, that's what I mean, there are a lot of networks, you can't master them without support,
-all this is where? Of course not in MT and not in the codebase, but in the NN package of course + friendly shell.
as well,
if someone can program, he knows enough maths to work with the NN package.
and vice versa,
only after the NN package the user can evaluate what is in CodeBase

 

I have NS2 installed, I haven't used it yet, I need more information to study.... it is possible to write a network in MQL, but it will work slowly, Kadet recently posted

i've got my doubts about the net... but Better showed very good results, it means something can be survived with correct use

 
Korey >> :

to Andrey4-min

Well, I mean the same thing, there are a lot of networks, without support you can not master,
-You can get it all from where? Of course, not in MT and not in the codbase, but in the NN package of course + friendly shell.
as well,
if someone can program, he knows enough maths to work with the NN package.
and vice versa,
only after the NN pack the user can evaluate the code in CodeBase

Where can I get the NN package?

 
if you take apart the NS2 examples, yes + try to generate a forecast two bars ahead, it's quick, and then unload the grid into MQL-4,
and that's how you get the training.
And then it will be clear, for example, why Reshetov uses a perseptron without scaling and how this perseptron differs from the network.
 
Yes, the area is really extensive, discussion forums, books, articles found with your help, which is thanks, honestly ... my head is spinning, I do not know what to grasp at, without reinventing the wheel all the same, so the theme of building a simple Expert Advisor here potderzhivayu!
 
NeuroShell, Statistica 8, Mathcad/Mathlab +NN
NeuroShell 2 is better for learning (and for a good job).
torrent rules
123456789...19
New comment