Discussion of article "Neural networks made easy (Part 32): Distributed Q-Learning"

New comment
 

New article Neural networks made easy (Part 32): Distributed Q-Learning has been published:

We got acquainted with the Q-learning method in one of the earlier articles within this series. This method averages rewards for each action. Two works were presented in 2017, which show greater success when studying the reward distribution function. Let's consider the possibility of using such technology to solve our problems.

While the testing EA was running in the MetaTrader 5 strategy tester for two weeks, trading based on the model signals, it generated a profit of about $20. All operations had a minimum lot. The below graph demonstrates a clear upward trend in the balance value.

Model testing in the strategy tester

Testing a distributed Q-learning model

Trading operations statistics shows that almost 56% of operations were profitable. However, please note that the EA is intended solely for testing the model in the strategy tester and is not suitable for real trading in the financial markets.

Author: Dmitriy Gizlyk

 
How many Era's do you train ?
 
try compile all yours EA from chapter 28 till 35 and still get same error..  can you help me sir..?
 
tey to use VAE file from chapter 22.. no error in compile but when i attach train EA nothing happen.. i missed something? 
 

Isn't this always zero:


I guess it should have been (Vmax-Vmin)/N?

 
Carl Schreiber # :

Isn't this always zero:


I guess it should have been (Vmax-Vmin)/N?

Hello, you are right.

 

What about #include "..\Unsupervised\AE\VAE.mqh" ?
There is simple manual how to use this stuff, if VAE.mqh will appear?

*edit* Here is the VAE.mqh https://www.mql5.com/en/articles/11245 it's on Part 22

Neural networks made easy (Part 22): Unsupervised learning of recurrent models
Neural networks made easy (Part 22): Unsupervised learning of recurrent models
  • www.mql5.com
We continue to study unsupervised learning algorithms. This time I suggest that we discuss the features of autoencoders when applied to recurrent model training.
 
IcHiAT #:

E quanto a #include "..\Unsupervised\AE\VAE.mqh" ?
Existe um manual simples de como usar essas coisas, se VAE.mqh aparecer?

*editar* Aqui está o VAE.mqh https://www.mql5.com/en/articles/11245 está na Parte 22

Yes, but I still get an error as below. Did you find this problem?


'MathRandomNormal' - undeclared identifier VAE.mqh 92 8

',' - unexpected token VAE.mqh 92 26

'0' - some operator expected VAE.mqh 92 25

'(' - unbalanced left parenthesis VAE.mqh 92 6

',' - unexpected token VAE.mqh 92 29

expression has no effect VAE.mqh 92 28

',' - unexpected token VAE.mqh 92 48

')' - unexpected token VAE.mqh 92 56

expression has no effect VAE.mqh 92 50

')' - unexpected token VAE.mqh 92 57


New comment