Discussion of article "Neural Networks: From Theory to Practice" - page 7

 
Yedelkin:
Wow... So we're not gonna be able to create a self-learning programme any time soon. Too bad. I'll remain a dummy :)
You can take ready-made developments, they have already been mentioned on the quad forum and here. For example, there is such a header file NeuroNet.mqh, with the help of which you can make an Expert Advisor that learns online and by itself (without the built-in optimiser). Only to start working with such classes, you need to read a couple of articles. The articles have already been published on the quaternary site, you can read them there, not wait for a sequel to be published here.
mt5fann
mt5fann
  • code.google.com
Forex Projects
 
marketeer: You can take ready-made developments, they have already been mentioned on the quad forum and here. For example, there is such a header file NeuroNet.mqh, with the help of which you can make an Expert Advisor that learns online and by itself (without the built-in optimiser). Only to start working with such classes, you need to read a couple of articles. There were already articles on the quadruple site, you can read them there, not wait for a sequel to be published here.
Thanks for the kick in the right direction!
 

Guys, is there any recommendations on choosing indicators, frames and pairs for the network? Maybe the author will share his experience?

I have noticed that the price spread has a certain pattern, maybe it can be used in training?

 
Алгоритм обратного распространения ошибки с регуляризацией на c#
Алгоритм обратного распространения ошибки с регуляризацией на c#
  • habrahabr.ru
Привет. Я хочу продолжить тему реализации методов машинного обучения на c#, и в этой статье я расскажу про алгоритм обратного распространения ошибки для обучения нейронной сети прямого распространения, а также приведу его реализацию на языке C#. Особенность данной реализации в том, что реализация алгоритма абстрагирована от реализаций целевой...
 

Why do such errors occur when testing the Expert Advisor from the article?

2012.11.14 20:02:08 2012.10.09 22:00:00 CTrade::OrderSend: instant sell 0.10 EURUSD at 1.2881 [invalid volume]

2012.11.14 20:02:08 2012.10.09 21:00:00 CTrade::OrderSend: instant buy 0.10 EURUSD at 1.2866 [invalid volume]

2012.11.14 20:02:08 2012.10.09 21:00:00 failed instant buy 0.10 EURUSD at 1.2866 [Invalid volume]

 
nskforward:

Why do such errors occur when testing the Expert Advisor from the article?

1. Let's start with which one? There are 2 Expert Advisors in the article: neuro-example.mq5 and macd-neuro-example.mq5.

2. Just downloaded both of them, ran them on the tester from this year, all ticks, EURUSD, MetaQuotes-Demo server - no problems found.

 
nskforward: Why do such errors occur when testing the Expert Advisor from the article?

2012.11.14 20:02:08 2012.10.09 22:00:00 CTrade::OrderSend: instant sell 0.10 EURUSD at 1.2881 [invalid volume]

2012.11.14 20:02:08 2012.10.09 21:00:00 CTrade::OrderSend: instant buy 0.10 EURUSD at 1.2866 [invalid volume]

2012.11.14 20:02:08 2012.10.09 21:00:00 failed instant buy 0.10 EURUSD at 1.2866 [Invalid volume]

What is the minimum allowable transaction volume on the account?
 
I apologise. Indeed, both Expert Advisors from the article work perfectly well. I just wanted to try them so badly that I registered a demo account with the first brokerage company that provides demo-access to MT5. It turned out that I can't open trades of less than 1 lot :(
 
nskforward: Indeed, both EAs from the article work perfectly.
No. "Excellent work" is when the Expert Advisor sends trade requests taking into account the available limitations for a particular account.
Документация по MQL5: Торговые функции / OrderSend
Документация по MQL5: Торговые функции / OrderSend
  • www.mql5.com
Торговые функции / OrderSend - Документация по MQL5
 
Yedelkin:
No. "Excellent performance" is when the EA sends trade requests taking into account the available limitations for a particular account.
Thanks for the idea. The article has been corrected.