Neural Network in Native MQL ?

New comment
 

Hi All:

I see lots of articles on this site on how to interface with MatLab, ENCOG and other such Neural Network engines using CSV, DDE, DLLs, and other such things, which is great. What I am wondering is if anyone has seen a neural net package written in native metaquates language?

I have been studying the ENCOG for C/C++ and I was thinking it would perhaps be more seamless and cut through some latency to write MQL classes which mirror the C++ functionality. I don't want to re-create the wheel if something like this has been done however. Is this a good idea? Is there any reason why translating the C++ to MQL5 wouldn't be desirable or workable? Has this already been done? Is it worth the effort?

Thanks!

DeWitt

 
SquareThree:


I have been studying the ENCOG for C/C++ and I was thinking it would perhaps be more seamless and cut through some latency to write MQL classes which mirror the C++ functionality. I don't want to re-create the wheel if something like this has been done however. Is this a good idea? Is there any reason why translating the C++ to MQL5 wouldn't be desirable or workable? Has this already been done? Is it worth the effort?


Good question, but my vision is that it's a hard work to create an up-to-date version, since such frameworks need to be refreshed all time.

By the way, solutions like this article (https://www.mql5.com/en/articles/252) are easier to update at each new version or system evolution.

Using MetaTrader 5 Indicators with ENCOG Machine Learning Framework for Timeseries Prediction
Using MetaTrader 5 Indicators with ENCOG Machine Learning Framework for Timeseries Prediction
  • 2011.04.25
  • investeo
  • www.mql5.com
This article presents connecting MetaTrader 5 to ENCOG - Advanced Neural Network and Machine Learning Framework. It contains description and implementation of a simple neural network indicator based on a standard technical indicators and an Expert Advisor based on a neural indicator. All source code, compiled binaries, DLLs and an exemplary trained network are attached to the article.
 
figurelli:

Good question, but my vision is that it's a hard work to create an up-to-date version, since such frameworks need to be refreshed all time.

By the way, solutions like this article (https://www.mql5.com/en/articles/252) are easier to update at each new version or system evolution.

 

Yes, I can see that updating the framework would be a real challenge. Thanks for your reply. DeWitt 

 
SquareThree:
Please don't reply inside the quote. Thank you.
 
SquareThree:
Thanks, you are welcome.
New comment
Reason: