Machine learning in trading: theory, models, practice and algo-trading - page 1427
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
NS is the same fit as the optimizer, even more so.
A philistine's position is based on a simple truth: you don't need to be a scientist to use electricity and wi-fi. So it is the same here: you - scientists, using analysis, which by its nature implies division into parts followed by synthesis, collecting from scraps of something supernatural, while we, nubs, are waiting on the sidelines when you attach an interface to your brainchild and show "where to poke")) I "created" a neural network by the instructions, trained it and "wrote" an Expert Advisor in MQL5 (I'll never learn any of them in my life). However, since it is considered a mauvais ton to reveal the secrets of successful algorithmic trading at MQL5, they usually write "The aim of this article is not to create a profitable trading system..." at the end of it, and leave room for maneuver. I'm waiting for another scientist who will tell me what to "press" next)))
In my opinion, this is the problem - there is no logic, which means complete freedom for the best approximation to the price history. For a stationary process this would be the right approach, but in our case it is not complete, and therefore not stationary. Use predictors with ideas - in my opinion this is more correct.
By the way, how many input neurons and layers with neurons on each layer in the network?
Thanks for the clarification.
This picture shows network with 3 layers with 20 neurons each. The maximum 10 layers of 100 neurons can be created in the program (it takes a long time to train, I tried to click it once)
No MOs or neural networks are useful in forex.
The NS is the same fitting as the optimizer, even more so.
In general, roughly speaking, this is the case. The optimizer, grid or "genetic", searches well for an optimum for a small number (a couple of tens) of relatively independent parameters, metrically distributed relatively homogeneously, when parameters are highly dependent and their distribution is far from homogeneous, this also applies when there are many of them (curse of dimensionality), stochastic search, let alone grid search, is not effective. For example, why can't we train a multilayer perseptron with a genetic algorithm? It can be done by reverse propagation by "fitting" hundreds or even thousands (all weights of all network neurons) parameters, which of course can "photograph" any multivariate noise, taking the art of retuning to a new level, comparing to dabbling with indicators.
Suitable. Very good for solving some classes of problems. Which ones, it is written in any book on NS.
In some tasks, of course, they are indispensable. For example in chess, or when recognizing fingerprints and different images, in standard movements, etc.
Here, here is the stumbling block, waiting for another scientist who will tell me what to "press")).
Pressing the buttons is no longer a problem. You don't even need to know anything anymore - just press the keys.
The whole question is to set a specific task for the Ministry of Defense. And it is unlikely anyone will do it for you or for you.
In some tasks, of course, they are indispensable. For example in chess, or when recognizing fingerprints and different images, in standard movements, etc.
Pressing buttons is no longer a problem. You don't need to know anything anymore, you just press the keys.
The whole question is to set a specific task for the MO. And it except for you and for you hardly anyone will do.
As far as I understand, the initial problem for the MO at Forex - is a figure recognition on the chart. And all the rest is attempts to pervert NS.)
And then I'll move on to the next article, and use NeuroSolution (or something like that) in it.
In some tasks, of course, they are indispensable. For example in chess or in the recognition of fingerprints and various images, in standard movements, etc.
Well, and on the market, too. As an example, learning logic. Instead of writing endless conditions yourself, put there NS - 15 minutes, and it's done. The question was how to put NS there).
As I understand it, the initial task for NS in forex is the recognition of figures on the chart. And all the rest - attempts of perversion over NS)))
If there is a rule of thumb like that the market does not recognize graphical "patterns", in this case the market does not remember them, does not "see" them, how many versions of "pettern correlators" that broke the pettern hypothesis down has already been presented.