Discussing the article: "Neural networks made easy (Part 45): Training state exploration skills" - page 2
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Cleared all tester logs and ran Research optimisation for the first 4 months of 2023 on EURUSD H1.
I ran it on real ticks:
Result: 4 samples in total, 2 in plus and 2 in minus:
Maybe I'm doing something wrong, optimising the wrong parameters or something wrong with my terminal? It's not clear... I'm trying to repeat your results as in the article...
The errors start at the very beginning.
The set and the optimisation result, as well as the agent and tester logs are attached in the Research.zip archive
1. I put full optimisation, not fast optimisation. This allows for a complete enumeration of the given parameters. And, accordingly, there will be more passes.
2. The fact that there are profitable and unprofitable passes when launching Research is normal. At the first run the neural network is initialised with random parameters. Adjustment of the model is carried out during training.
The problem is that you run "tester.ex5". It checks the quality of trained models, and you don't have them yet. First you need to run Research.mq5 to create a database of examples. Then StudyModel.mq5, which will train the autoencoder. The actor is trained in StudyActor.mq5 or StudyActor2.mq5 (different reward function. And only then tester.ex5 will work. Note, in the parameters of the latter you need to specify the actor model Act or Act2. Depends on the Expert Advisor used to study Actor.
Dmitry good day!
Can you tell me how to understand that the training progress is going at all? Do the percentages of error in reinforcement learning matter or do they look at the actual trading result of the network?
How many cycles did youstudy (StudyModel.mq5 -> StudyActor2.mq5 ) until you got an adequate result?
You indicated in the article that you initially collected a base of 50 runs. Did you make additional collections in the process of training? Did you supplement the initial base or delete and recreate it in the process of training?
Do you always use 100,000 iterations in each pass or do you change the number from pass to pass? What does it depend on?
I taught the network a lesson for 3 days, I did maybe 40-50 cycles. The result is like the screenshot. Sometimes it just gives a straight line (does not open or close trades). Sometimes it opens a lot of trades and does not close them. Only equity changes. I tried different examples base. I tried to create 50 examples and then make loops. I tried to create 96 examples and added another 96 examples every 10 cycles, and so on up to 500. The result is the same. How do I learn it? What am I doing wrong?
Good afternoon Dimitri!
Can you tell me how to understand that the progress of training is going at all? Do the percentages of error in reinforcement learning matter or do they look at the actual trading result of the network?
How many cycles did youstudy (StudyModel.mq5 -> StudyActor2.mq5 ) until you got an adequate result?
You indicated in the article that you initially collected a base of 50 runs. Did you make additional collections in the process of training? Did you supplement the initial base or delete and recreate it in the process of training?
Do you always use 100,000 iterations in each pass or do you change the number from pass to pass? What does it depend on?
I taught the network a lesson for 3 days, I did maybe 40-50 cycles. The result is like the screenshot. Sometimes it just gives a straight line (does not open or close trades). Sometimes it opens a lot of trades and does not close them. Only equity changes. I tried different examples base. I tried to create 50 examples and then make loops. I tried to create 96 examples and added another 96 examples every 10 cycles, and so on up to 500. The result is the same. How do I teach it? What am I doing wrong?
Same thing...
Spent a few days, but the result is the same.
How to teach it is unclear ...
I have not managed to get the result as in the article....