Using artificial intelligence at MTS - page 22
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
IMHO, it's not worth steaming about with nets :)
Learning NS is actually optimizing a function with a huge number of parameters (hundreds and thousands).
I don't know what to do to avoid overtraining in this case,
The only solution is to take a training sample of 1-100 million samples.
But there is no guarantee...
It occurs when you have a lot of optimization parameters and not enough data.
But this raises the question of network size. What a network can store depends upon its size and architecture. If you give too many samples to be trained that the network cannot remember, it will cause the overlearning effect - the network will stop recognizing what it knows.
http://www.osp.ru/os/1997/04/179189/
http://www.exponenta.ru/soft/Others/mvs/stud3/3.asp
Question for mathematicians:
Is the idea of applying a multivariate normal distribution of the parameters to be optimised equal to the principle of neural networks?
Please explain it clearly.
Explain the question.
EXPLAIN:
I now think that it is necessary to trade not with specific fitted parameters, but with the spectrum of each parameter in the system.
The easiest way is to put several identical EAs, but with different sets of parameters - in different ranges of the parameter spectrum.
Each of these Expert Advisors should be allocated a certain % of the deposit, but all of them should be equal to the deposit percentage value, when trading using only one Expert Advisor (without spectrum).
Then if on moving averages three Expert Advisors open three positions, respectively at the beginning of movement in the middle and at the end.
I cannot yet decide how to use this idea in one EA for testing.
I asked Posh about this problem but still no answer.
The task of multivariate normal distribution (Gaussian) and neural networks of aX+bY+...=Z type are the same (for trading), or I got confused and am confused in my head?
I've never heard of anyone getting positive results.
In other words, it's called "diversification by parameters".
In MT it's simple - put everything you have in the start function into another function with parameters,
and call it in a loop in the start function, giving you the set of parameters you need.
It shouldn't work, at least because of the principle of superposition of positions.
(independent systems add up aditively - i.e. if each one separately fails, then the sum will also fail).
It has nothing to do with neural networks.
The appearance of the Gaussian distribution or its derivatives (e.g. lognormal distribution)
is usually indicative of "market efficiency" (normal distribution is a consequence of the central limit theorem).
And there is nothing for any system to catch.
Only "market inefficiency" can be traded.
Now I think that it is necessary to trade not with specific fitted parameters, but with the spectrum of each parameter in the system.
The easiest way is to set several identical Expert Advisors, but with different set of parameters - in different ranges of parameter spectrum.
Each of these Expert Advisors should be allocated a certain % of the deposit, but all of them should be equal to the deposit percentage value, when trading using only one Expert Advisor (without spectrum).
Then if on moving averages three Expert Advisors open three positions, respectively at the beginning of movement in the middle and at the end.
But I do not know how to put this idea into one Expert Advisor for testing yet.
With 'parameter diversification' the test results will definitely be worse than with over-optimised parameters.
But system survivability on new data should increase.
I will try with different MagikNumber in one Expert Advisor with output of custom evaluation results of each position in the "spectrum" to a file.
So, let me ask you a question.
What have you got?
With 'parameter diversification' the test results will definitely be worse than with over-optimised parameters.
But system survivability on new data should increase.
I will try with different MagikNumber in one Expert Advisor with output of custom evaluation results of each position in the "spectrum" to a file.
So, let me ask you a question.
What have you got?
The Expert Advisor template is VERY EASY to insert an unlimited number of strategies that work independently of each other.
So I have a universal Expert Advisor for all occasions (as probably everyone here already does).
But we are not talking about different strategies working at the same time, we are talking about one strategy working simultaneously with diversified parameters.
I have a universal Expert Advisor for all occasions (as probably everyone here already has one).
But we are not talking about simultaneous operation of different strategies, but about simultaneous operation of one strategy with diversified parametres.
I have not noticed the universality of your Expert Advisor, judging by...
But I have not yet figured out how to put this idea into one Expert Advisor for testing.
And the work of one strategy, with, as you write, diversified parameters, is even easier to implement than the work of several strategies.
If you have any idea, put it out there!
And everybody can blurt out an idea with their tongue.
And I have already put the idea into my expert.
It all worked out.
I'm thinking how best to visualize and comprehend.
How to filter orders in a chart after the test with a certain MagikNum?
Who knows?