Neural Network Trading, Serious people only!

Could you describe an installing guide for your indicator set and mt4 plugin?
i only get this by attaching NN Delta :
(MT4 Build 510)

.. to get your result:

I like NEUROSHELL -are there any trader feeds which don't use fibonacci to create their signal entries

I write this for those interested on what to search to apply RF learning to trading. The easiest RF learning algo would consist on just calculating returns based upon OHLCV data for instance.

Implementing the RFL agent, policy, value function, environment, actions and so on is only one part. Maybe, the definition of the environment may be challenging, but it is better to work bottom up.

But the key thing and this is why people usualy fails to forecast correctly using IA and Machine learning is that they focus only in one between exploration and exploitation. But not in both. Of course I am only talking about anyone able to distinguish among linear and non-linear systems, dependencies, experiment modelling, and so on.

When building a RFL system you must set a parameter for exploration with random actions.

Then, when you want to analyze your argmax to take a decision for an action, there are two types of causistics:
1) Deterministic (not good for forex).
2) Stochastic (the one that should prevail).

And there is when neural networks or any other prediction system should be added to the RFL system. You have to fill in your adjacency matrix with state values.

So the trick is to mix one good RFL model linked to the best possible predictor. Anyone can test results using MLPs, LSTMs, CNs deep learning models or anything else; like support vector machines, or even a simple logisitc regression.

But all this is quite abstract if you are not able to manually be profitable at forex, and for this reason I think that in order to improve our models we need to make an extra effort to learn to trade, because that will guide us directly for a better environment definition and actions configuration.

Metatrader can import native dlls (32 bit). Or we can use also bridges suitable for Python or C# for instance. I prefer the c++ way because I think it is more stable.

But sometimes simplicity is better than perfection.

Machine learning and RFL is not suitable for anyone. For instance, I am currently mentoring a matematician on how to design ML systems for time series and it is not a trivial issue even for him. Not to say the people I for instance met in my Data Science master, all engineers, matematicians, statisticians, etc...

But following the main guidelines I have posted I think a good design plan may be designed to start with RFL.

Hello. I advise you to use a neural network, recurrent

Holy Grail Deep Neural Network recurrent: excellent This expert advisor is available for you to enjoy, on MQL5.
4 H.

its amazing,but did you laod it on your real account?

Hello. Thank you for appreciating the work. Of course I have it in real accounts, from Fx Pro and Dukascopy bank brokers, and so far this month they have won 100% of the operations, I am using the pairs: EurUsd, and UsdJpy, the least manipulable, in my opinion.I think it is on its way to being a real money-making machine, not inflationary

Hello Farmabiog,
Can you tell me more about your system ? What type of inputs you used ? RNN Network ? preprocessing partiuclar ?
Regards.