"New Neural" is an Open Source neural network engine project for the MetaTrader 5 platform. - 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
Maybe we should bring in developers of neural systems who are already working in the field of trading as experts?
Call them from other sites.
Maybe as experts to involve the developers of neural systems who are already working in the field of trading?
Call them from other sites.
Are you talking about paid influencers or
Just about informing a wider audience in the hope of generating interest?
Are you talking about paid influencers or
just about informing a wider audience in hopes of generating interest?
For us, the task of developing the engine has already moved to the implementation stage. Today we had an internal discussion about the idea and moved on to preparing the infrastructure.
Now we need experts who will help in word and deed.
For us, the task of developing the engine has already moved to the stage of implementation. Today we had an internal discussion about the idea and moved on to preparing the infrastructure.
Now we need experts who will help us in word and deed.
Well, then you need to publish what you have.
There will be something to discuss.
But if there is no reaction, then we can scratch our heads.
But if there is no reaction, then you can scratch your head.
Ahem (modestly) about the reaction -- there are already 3 libs for neurons.
One has more than 10 nets. I've worked with Kohonen nets, MLP, recirculation nets, Hopfield nets... ,
the second one is a general case of MLP + Jordana-Elman network -- i.e. any topology (directed graph) with possibility to loop back any layers,
the third is an implementation of Echo Network, my favorite :) .
It was a long time ago really (except echo network), but it is possible to remember. Did not work with probabilistic models. Not familiar with recent improvements of gradient descent method and hybrid methods.
Ahem (modestly) about the reaction -- there are already 3 libs for neurons.
It was a long time ago really (except for the echo network), but you can remember. Haven't worked with probabilistic models. Not familiar with recent improvements to the gradient descent method and hybrid methods.Ok. Can all kinds of network topologies be provided as an overview? That is, the more the better, but that it be expressed as a structure or diagrams?
So that we can define the models to be implemented and thus start designing abstract general base classes.
the task, by the way, is not only to make some set of networks as a result, but also to lay down in classes the possibility to extend them to other particular network topologies.
Before you start something, prepare an accessible and understandable theory and practice for the people, and then start to create something.
This way, you can explain to future makers what is the advantage of an NS written only in MQL5 and what is the disadvantage. I don't think anyone would want to bother creating a package, which will work as an emulation in MT5, knowing that programs written in emulated languages are slower to work than in higher-level languages.
My advice. If everyone is eager to write anyway, create a mathematical model in a DLL, and prepare the data in MQL5. Implement each type of NS in separate DLL. For example, the Kohonen neural network will be in the file module_kohhonen.dll, and the Hopfield network module_hopfield.dll. Use the languages as intended and don't reinvent the wheel.
And a graphical interface is planned. For clarity, the ability to feel the structure of the network, etc.
Ok. Can all kinds of network topologies be provided as an overview?
OK, I'm interested in 4 networks that have been implemented
1. Kohonen networks, including SOM. Good to use for cluster partitioning where it is not clear what to look for. I think the topology is well known: vector as input, vector as output or otherwise grouped outputs. Learning can be with or without a teacher.
2. MLP , in its most general form, i.e. with an arbitrary set of layers organized as a graph with the presence of feedbacks. Used very widely.
3. Recirculation network. To be honest, I have never seen a normal working nonlinear implementation. Used for data compression and principal component extraction (PCA). In its simplest linear form, it is represented as a linear two-layer network, in which the signal can be distributed from both sides (or three-layer in expanded form).
4.Echo-net. Similar in principle to MLP, applied there. But it is absolutely different in organization and has clearly specified time of training (well, it always gives global minimum, unlike).
5. PNN -- I haven't used it, I don't know how. But I think there are people who can do it.
6. Models for fuzzy logic (not to be confused with probabilistic networks). Haven't implemented them. But could be useful. If anyone can find information, throw plz. Almost all models have Japanese authorship. Almost all of them are build by hand, but if it were possible to automate topology building by logic expression(if I remember correctly), it would be really cool.
_______________
Suggest other models.
All networks are representable as input--black box--output
You probably can't interface all the networks completely, though you should try.