Hi, I check the file... actually, I do not know where to place all those file. Which one is expert or indior, and another file, what's that? Are those pro or indicaator?
Publi |
Hi, I check the file... actually, I do not know where to place all those file. Which one is expert or indior, and another file, what's that? Are those pro or indicaator?
I'd love to assist as I was just considering doing some NN stuff for MetaTrader myself. I think the first thing you have to perform neuro if we are going to make this a collaborative project is to go through and clean up the source code and comment everything correctly. A lot of the variable names are very vague and the code has been strewn all about the file so leaping back and forth to trace everything is a pain, especially for somebody like me that doesn't have experience using these NN libraries. Do you think that could go through and comment all the code completely and wash this up to look nicer so we can all begin from a fantastic codebase? If we are going to develop something to become open source we need to be sure excellent coding standards are followed closely from the start.
This is the updated zip - added FannDoubleMT.dll (forgot to add in the prior post), and all copyright and proprietary material has been removed.
https://www.forexycfds.com/attachmen...1646716661.zip
Onu, Yes I am on the Neuraltraders staff - but this particular EA - NextClose will always be in the open source domain, perhaps we can also rope in a couple of very good people from sourceforge etc.. Collaborative effort grows everyone - even companies such as sunlight oracle develop software with open source gifts - so copyrights aren't an issue, anyways I will get rid of all neuraltrader associated stuff. 7Bit, Thanks for your feedback, I Believe we ought to benchmark with both 3 concealed layers and Might be look at cascade trading to come up with optimal number of ns but I believe we all will concur rather than making it flexible (multi pair) allows concentrate on a single pair and EURCHF using its relatively low disperse and rangebound nature will be a good candidate - but I am open to suggestions
IIRC you won't ever need more than 2 hidden layers since 2 layers are already sufficient to approximate any purpose you could approximate together with 3 or 3 layers. I can't remember precisely where I read this but IIRC that has been mathematically proven .Iniciado por ;
The copyright of the EA belongs to NeuralTraders.info U are the guy behind this membership item? Come on, we ai not here to develop your EAs, if that is the case, or to offend other individuals copyrights!
Http://www.myfxbook.com/community/su...gh-ai-/19892,1Alguien parece estar en una gira de promoción ...
---------------------------------------------
http://www.myfxbook.com/members/pc8multifx
Thanks Ronald to your curiosity and Onu and tulipeverige to your offer to check. I'm not a big fan of reinventing the wheel chose to utilize the excellent and very old FANN library, and then a person did the grunt work of hammering it to mql - Fann2mql undertaking, I however had to modify Fann2mql source code to permit for 3 hidden layers (by default allows for just two ). In another project on SP futures we found 3 layers produced smaller mse than two although it took more to train - more than 3 layers didn't enhance mse by much. Instructions: Copy the dlls to experts/libraries, ntann and ntcommons into includes.
https://www.forexycfds.com/attachmen...1570302676.zip
hi neuro, count me in! Please post more details, that we can start with the developement thanks, -ba