How to form the input values for the NS correctly. - page 14

New comment
 

YuraZ писал (а) >> в своем варианте я жестко сравнивал выход после обучения с реальным выходом на тестовом патерне, Вы толкько ошибку Ваш вариант пробую - так и не удалось еще ни разу обучить! в моем варианте обучение проходило быстрее

I've learned everything fine. The SetTeachPattern function sets the test input and output similar to yours.

The expression

MathAbs(Target[i]-ol_Out[i])>0.01

and yours

         if (OUT_PATTERNS[pat,0]==0.00  && OUT_PATTERNS[pat,1]==0.00  &&  OUT_PATTERNS[pat,2]==1.00 )
           {
            if(ol_a[0]<= 0.01 && ol_a[1]<= 0.01 && ol_a[2]>= 0.9)
            //
           }

are identical.

 

Has anyone seen this? A trading system with GA! Thanks to the author.

Files:
0_ema_neuro_ga.mq4  14 kb
 
StatBars писал (а) >>

Has anyone seen this? A trading system with GA! Thanks to the author.

and even tried to write an EA... Too bad it doesn't take profits...

But great, there is also GA implemented in MQL4, it's from KLOTA site


http://fxreal.ru/forums/ KLOT NEUROSETS resource

http://fxreal.ru/forums/topic.php?forum=2&topic=9 KLOT Resource Genetic Algorithms


The method of feeding the difference of the dashes to the input

 
YuraZ писал (а) >>

and even tried to write to an advisor... Too bad it doesn't take profits...

but great, there's also GA in MQL4, it's from KLOTA's site

Yes, it is. I meant not to look for entry points by myself but to provide this GA. But one point per position is not enough, so I propose to do like this - GA opened Buy - all points lying nearby and not exceeding Y points from the initial point are sent to the training sample. If GA gave a drawdown as it can be seen in the figure all points lying below also should go to SP.

 
StatBars писал (а) >>

Yes yes. My point is not to look for entry points on your own, but to provide them to the GA.

The GA-based optimiser also has a significant disadvantage that many people don't consider. The GA may not find the right solution if it does not give the "right" range for finding parameters. If it (this range) is too small or too large. That's why you have to be very careful with GA.

And its significant advantage is that it's faster than usual optimizer which tries all existing variants.

 
StatBars писал (а) >>

Yes yes. I mean not to search for entry points by myself, but to let the GA do it. But one point per position is not enough, so I propose to do this - GA opened Buy - together with this point all points lying nearby and not exceeding Y points from the starting point are sent to the training sample. If GA has caused a drawdown, as it can be seen in the picture, all points lying below should also go to SP.

from this point! yes points are very good

if there is a divergence or covvergence at this point, the point should probably be considered 100%

---

but unfortunately this indicator does not always look for the best points

---

there are a lot of methods to find reversals --- you can see them with your eyes perfectly of course

but the software is not always successful - I mean on the history

this indicator is tied to a specific EMA...

there are other indicators that are much better at "heroic" on the history

 
YuraZ писал (а) >>

from this point of view! yes points are very good

if there is a divergence or covvergence at this point, the point should probably be considered as 100%

---

but unfortunately this indicator does not always look for the best points

---

there are a lot of methods to find reversals --- you can see them with your eyes perfectly of course

but the software does not always work well - I mean on the history

this indicator is tied to a specific EMA...

there are other indicators that are much better "heroic" on the history

Show me...

 
StatBars писал (а) >>

Show me...

the first normal zigzag!

with convergence on several timeframes in one point

- there was an indicator based on it somewhere on the forum!

it is a beauty! - BUT ON THE STORY

I can't remember the name, but I can find it.

 
YuraZ писал (а) >>

>> the first regular zigzag!

Nah, zigzag's no good, got anything else?
 
YuraZ писал (а) >>

the first normal zigzag!

with convergence on several timeframes in one point

- there was an indicator based on it somewhere on the forum!

it is great! - BUT ON HISTORY

I can't remember the name - but I can find it.

Is it? At least it looks like it... It's a three-period indicator...

I don't think it fits.

Files:
3_level_zz_semafor.mq4  8 kb
1...789101112131415161718192021...31
New comment