Discussion of article "Third Generation Neural Networks: Deep Networks" - page 9
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
Hi Vladimir,
My name is Krzysztof (krzysiaczek99) and I run similar thread about 3rd generation NN on trade2win since 2010. In my implementation I use MATLAB and matlab toolboxes and also WEKA to have access to different algorithms.
From my experience deep nets did't do well with financial TS, much more efficient was e.g SVM. So did you fix all files in your implementation as I would like to try ??? Is backtest possible ??
Krzysztof
Hi,
1. The revised expert will present in a new article that ends. It is a deep neural network with RBM.
2. On the MT4 expert test I failed.
3. The benefits of the other models of machine learning to deep networks - a very controversial statement. In my experience it is not.
Best regards
Vladimir
It would be very interesting if you would be able to show your AI-approach compared to a classic EA-Optimization.
Take e.g. a free successful EA (like https://www.mql5.com/en/code/913) EA_MARSI - expert for MetaTrader 5, Description: The Expert Advisor is based on EMA_RSI_VA indicator.
It seems to bee a quite simple EA how would you code this classic EA-approach to fit into your AI idea?
You can even try to add - just to show how it is done - some other indicators (e.g. CCI, BollingerBands, ZigZag) as a try to improve this EA - no matter whether this idea might be successful or not.
It would be very interesting if you would be able to show your AI-approach compared to a classic EA-Optimization.
Take e.g. a free successful EA (like https://www.mql5.com/en/code/913) EA_MARSI - expert for MetaTrader 5, Description: The Expert Advisor is based on EMA_RSI_VA indicator.
It seems to bee a quite simple EA how would you code this classic EA-approach to fit into your AI idea?
You can even try to add - just to show how it is done - some other indicators (e.g. CCI, BollingerBands, ZigZag) as a try to improve this EA - no matter whether this idea might be successful or not.
Hi,
You can, but why?
If there is free time can be.
Best regards
Hi,
You can, but why?
If there is free time can be.
Best regards
1) It would help a lot those people like me who have a traditional approach to develop a trading idea based on some indicators.
2) We would have a traditional EA and your EA with the same basic approach to compare (additional or less work, ...)
3) It would prove how helpful your way would be.
1) It would help a lot those people like me who have a traditional approach to develop a trading idea based on some indicators.
2) We would have a traditional EA and your EA with the same basic approach to compare (additional or less work, ...)
3) It would prove how helpful your way would be.
Hi,
No task is to prove that this is the most effective way. The challenge to show another way to other, more features. And everyone has to choose their own, comparing them, if necessary, personally. Using machine learning in trading requires a fairly high level of knowledge and experience in the trade.
It all depends on the experience, knowledge and goals. Someone trading without indicators, and it is quite satisfied. Someone found some good work indicators and it suits him, too. Who is this just is not enough, because it is more deeply understand the market.
Sorry. It's hard to explain to You, "language barrier" prevents.
Hi,
No task is to prove that this is the most effective way. The challenge to show another way to other, more features. And everyone has to choose their own, comparing them, if necessary, personally. Using machine learning in trading requires a fairly high level of knowledge and experience in the trade.
It all depends on the experience, knowledge and goals. Someone trading without indicators, and it is quite satisfied. Someone found some good work indicators and it suits him, too. Who is this just is not enough, because it is more deeply understand the market.
Sorry. It's hard to explain to You, "language barrier" prevents.
:(
This is quite disappointing.
I thought you would be interested to perform a little competition to prove the usefulness of your approach.
You know most programmer describe their program what and how it calculates this and that - but hardly how the user will benefit from their program.
Don't forget that "..the most effective way." at least for me includes the amount of code I have to understand and to write and to optimize.
I (may be others too) really would be glad to get a feeling about all that - beside the naked performance.
Anyway thanks for your effort and patience.
Calli
:(
This is quite disappointing.
I thought you would be interested to perform a little competition to prove the usefulness of your approach.
You know most programmer describe their program what and how it calculates this and that - but hardly how the user will benefit from their program.
Don't forget that "..the most effective way." at least for me includes the amount of code I have to understand and to write and to optimize.
I (may be others too) really would be glad to get a feeling about all that - beside the naked performance.
Anyway thanks for your effort and patience.
Calli
to do such comparative test you have to be sure that any of this methods has really predictive power otherwise it is comparision of random results with unknown variance depending of the setup and chosen data so outcome is random also. Are you sure if any of those methods has really predictive power ?? Than it would require a lot of testing against a lot of symbols and even after this you would not be sure if one method is better than other as the variance distribution of the results is unknown and results it will just depend of test setup and chosen data.
Krzysztof
Hi,
1. The revised expert will present in a new article that ends. It is a deep neural network with RBM.
2. On the MT4 expert test I failed.
3. The benefits of the other models of machine learning to deep networks - a very controversial statement. In my experience it is not.
Best regards
Vladimir
So when you plan to release your new article ??
Krzysztof
to do such comparative test you have to be sure that any of this methods has really predictive power otherwise it is comparision of random results with unknown variance depending of the setup and chosen data so outcome is random also. Are you sure if any of those methods has really predictive power ?? Than it would require a lot of testing against a lot of symbols and even after this you would not be sure if one method is better than other as the variance distribution of the results is unknown and results it will just depend of test setup and chosen data.
Krzysztof
Thanks for your patience with me!
" ...Are you sure if any of those methods has really predictive power ??"
I don't want you to develop a perfect system with the absolute perfect indicators predicting the future!
We all know indicators only compress the existing chart of elapsed quotes into numbers and we attribute these numbers to the future. Right or wrong will tell us the market.
"... so outcome is random also."
This I don't understand. :(
In order to get comparable results I suggested an existing EA (MARSI) which is based on a known indicator EMA_RSI and asked you to use the same strategy.
If you backtest your EA and the existing MASI on the same historic data the final results of both EA (yours and the MARSI-EA) are comparable with each other - but still valid is: good or bad will tell us the market.
calli