The Ultimate AI EA Project - page 4

 
Thiago Martins:

Hello Nelson,

Thank you so much for your fast reply and help! I really appreciate it. You rock!

I'll give it a try tomorrow evening. Time to recharge now...  ^_^

Thanks again buddy!

Welcome✔

 

Please find attached a WORKING fully connected Neural Network Model with a sample script to help you test functionalities in a few clicks.

More work is in progress.

edit:

1000 was the default number of epochs in the provided files. However, learning converges at less than 250 iterations.


Among the tasks in progres, I'm aiming at creating an auto-configuration module which will select the most efficient network structure, and also simplify data input and training procedures.

I intend to use a single data container to send and retrieve predictions from the network for efficiency purposes.

Happy trading.

Documentation on MQL5: Constants, Enumerations and Structures / Named Constants / Predefined Macro Substitutions
Documentation on MQL5: Constants, Enumerations and Structures / Named Constants / Predefined Macro Substitutions
  • www.mql5.com
//| Expert initialization function                                   | //| Expert deinitialization function                                 | //| Expert tick function                                             | //| test1                                                            |...
 

Some changes have been made for ease of use of the script.

Check the attached change-log.

Files:
 

Hello


I do have some libraries I've developed myself. The most interesting things  I have related to AI are a Convolutional Neural Networks (which I'm using for 1D convolution of timeseries) and a Generic Q Learning Algorithm for reinforcement learning (which can be used coupled with neural networks or any function that is an approximation of a q-table) both in programmed in mql5.  I do have a git repository with all that content.

About "Shared projects", I personally love to use git in order to track my personal AI project.

If you guys have any interest, let me know.

 
Rogério Santana:

Hello


I do have some libraries I've developed myself. The most interesting things  I have related to AI are a Convolutional Neural Networks (which I'm using for 1D convolution of timeseries) and a Generic Q Learning Algorithm for reinforcement learning (which can be used coupled with neural networks or any function that is an approximation of a q-table) both in programmed in mql5.  I do have a git repository with all that content.

About "Shared projects", I personally love to use git in order to track my personal AI project.

If you guys have any interest, let me know.

Hi. Please tell us more about your q-learning algorithm. I've been doing a research in that domain with intent to create a custom rainforcement learning algorithm bundled with neural networks. Is it ready yet or still under development?
 

https://www.mql5.com/en/code/30815


Happy weekend

Easy Neural Network
Easy Neural Network
  • www.mql5.com
The provided test script has included comments which are easy to follow along. You can save the returned network configuration in a file then load it in different sessions...
 
Nelson Wanyama:
Hi. Please tell us more about your q-learning algorithm. I've been doing a research in that domain with intent to create a custom rainforcement learning algorithm bundled with neural networks. Is it ready yet or still under development?
Hi, I have it already developed. I do have an example of it working with the cartpole game (all developed in mql5). Unfortunately, it did not work with the indicators I was using, but the library is there. Currently, I'm working with supervised learning (which is faster) and once I find a set of variables that fit together and make a great model, I maybe try to assemble it in q-learning again. Anyways, I do have the q-learning library developed. I'll message you more intel 
 
Rogério Santana:
Hi, I have it already developed. I do have an example of it working with the cartpole game (all developed in mql5). Unfortunately, it did not work with the indicators I was using, but the library is there. Currently, I'm working with supervised learning (which is faster) and once I find a set of variables that fit together and make a great model, I maybe try to assemble it in q-learning again. Anyways, I do have the q-learning library developed. I'll message you more intel 
Amazing! Keep pushing
 
Hi guys, I am currently studying ML for my algorithms and I have found a very good article (not known here I think) about this subject.
I am happy to share it with you :
https://www.mql5.com/en/blogs/post/724245
Of course this isn't a graal but I think it can help beginners to visualize the multi layer perceptron (not in depth of course you will need a lot more mathematics).
So like Nelson said keep pushing and good luck in your work :).
See U mates.
Programming a Deep Neural Network from Scratch using MQL Language
Programming a Deep Neural Network from Scratch using MQL Language
  • 2019.01.27
  • www.mql5.com
Since machine learning has recently gained popularity, many have heard about Deep Learning and desire to know how to apply it in the MQL language. I have seen simple implementations of artificial neurons with activation functions, but nothing that implements a real Deep Neural Network. In this article, I will introduce to you a Deep Neural...
 

Hello. 

In my research, I constantly tried to figure out why existing trading algorithms perform, let's say not so well. And I found the answer to be, 'NOISE'.

This is an additional barrier that I thought I should cross, as part of the Ultimate AI Trading Algorithm. To achieve this, I have decided to construct renko charts using custom symbols, which will help to filter noise from the price time-series, which has been introduced by time-based scales. 

I like to continue hoping that more people will join efforts to make this project a  success. 

Happy trading.