Machine learning in trading: theory, models, practice and algo-trading - page 256

 
Top2n:

Each array[][] is a set of single-quality information, that is, for each input, a separate array[][]. I want to feed a lot of arrays, I'm ready for now 4, I'll create more in the plans, each array describes the state of the price, it turns out from different angles, something like that.

Each array contains 1000 rows and 1000 columns, well, in general, I have a three-dimensional, it turns out that the K-dimension is the new two-dimensional
I do not understand what you want to do and therefore will not say anything on this subject, but I say another thing, judging by the number of inputs, you can start to study methods of compressing information such as "PCA" or others, you will need them
 

Usually networks work with two-dimensional matrices. You could combine all the matrices into one broad matrix with 2000 (3000 or 4000) columns and 1000 rows, and then train the neural network as usual. You can train neuronics in R with rattle, it's a special program with a visual interface for data analysis, where you can use buttons and menus to process data and teach the model, and then in the "log" tab look at the generated R script for all these operations, and then you can just change and run the resulting code in the R console itself. That is you do all operations with your mouse and then you can watch the generated R script which does the same thing, it's convenient to study the possibilities of the language.
https://www.mql5.com/ru/articles/1165 - there is a lot and good about rattle.

There's also the mxnet package for R, it's a deep learning neuron, and trains on 4-dimensional arrays, even more measurements than you need
There are installation instructions (Installlation) and some examples in vignettes and demo folders -https://github.com/dmlc/mxnet/tree/master/R-package

 
mytarmailS:
I do not understand what you want to do, so I will not say anything about it, but I say something else judging by the number of inputs, you can start to study methods of compressing information such as "PCA" or others, you will need them

The logic of your question is probably deeper than my explanation.

I want to feed a couple of two-dimensional arrays into the network.

The arrays are the same along the axes, i.e. the columns are graded by period and the rows by bar.

 
Dr.Trader:

Usually networks work with two-dimensional matrices. You could combine all the matrices into one broad matrix with 2000 (3000 or 4000) columns and 1000 rows, and then train the neural network as usual. You can train neuronics in R with rattle, it's a special program with a visual interface for data analysis, where you can use buttons and menus to process data and teach the model, and then in the "log" tab you can see the generated R script for all these operations, and then you can just change and run the resulting code in the R console itself. That is you do all operations with your mouse and then you can watch the generated R script which does the same thing, it's convenient to study the possibilities of the language.
https://www.mql5.com/ru/articles/1165 - there is a lot and good about rattle.

There's also the mxnet package for R, it's a deep learning neuron, and trains on 4-dimensional arrays, even more measurements than you need
There's an installation guide (Installlation) and some examples in the vignettes and demo folders -https://github.com/dmlc/mxnet/tree/master/R-package

Thanks!
 
Top2n:

The logic of your question is probably deeper than my explanation.

I want to feed a couple of two-dimensional arrays into the network.

The arrays on the axes are the same, that is, the columns are graded by period, and the rows by bar.

I don't have any deep logic)))

You just have a lot of inputs into the network, it will take a long time to learn, you will come to the point where you have to reduce the number of inputs, that's why I advised you to study compression methods

 
Top2n:
Thanks!
Dr.Trader:

Usually networks work with two-dimensional matrices. You could combine all the matrices into one broad matrix with 2000 (3000 or 4000) columns and 1000 rows, and then train the neural network as usual. You can train neuronics in R with rattle, it's a special program with a visual interface for data analysis, where you can use buttons and menus to process data and teach the model, and then in the "log" tab look at the generated R script for all these operations, and then you can just change and run the resulting code in the R console itself. That is you do all operations with your mouse and then you can watch the generated R script which does the same thing, it's convenient to study the possibilities of the language.
https://www.mql5.com/ru/articles/1165 - there is a lot and good about rattle.

There's also the mxnet package for R, it's a deep learning neuron, and trains on 4-dimensional arrays, even more measurements than you need
There's an installation guide (Installlation) and some examples in the vignettes and demo folders -https://github.com/dmlc/mxnet/tree/master/R-package

Would you happen to tell me if I did everything according to the instructionshttps://www.mql5.com/ru/articles/1165

  • Load R;
  • loading the Rattle library;
  • File/Workspace tab;
  • look for the TC.RData file on the disk and load it.

And it's silent.


Случайные леса предсказывают тренды
Случайные леса предсказывают тренды
  • 2014.09.29
  • СанСаныч Фоменко
  • www.mql5.com
В статье описано использование пакета Rattle для автоматического поиска паттернов, способных предсказывать "лонги" и "шорты" для валютных пар рынка Форекс. Статья будет полезна как новичкам, так и опытным трейдерам.
Files:
yypb0b9hyu.jpg  197 kb
 
Top2n:

Would you happen to tell me if I did everything according to the instructionshttps://www.mql5.com/ru/articles/1165

  • Load R;
  • loading the Rattle library;
  • File/Workspace tab;
  • look for the TC.RData file on the disk and load it.

It doesn't say anything.


Try just typing

library(rattle)
 
mytarmailS:

Try just typing

library(rattle)

Man, I don't want to bother you with this nonsense, but if you know, maybe something standard is missing.

*** added

All simply entered rattle()

Files:
l9vzf3tyb7.jpg  142 kb
 
There is also Neurosolutions. You can generate dlls and plug them into the Expert Advisor
 
Top2n:

Man, I don't want to bother you with this nonsense, but if you know, maybe something standard is missing

*** added

It's easy to enter rattle()

You can't see rattle() on the picture.

After entering this command, the rattle window itself should appear, as described in the article

TC.RData (as I understand from the article) must be loaded into rattle().