Question on neural network programming - page 3

New comment
 
to VBAG

Who told you that нейросети is for predictions?

It's always nice to talk to a knowledgeable person. That's some incredible "brains" behind those four letters. Tell me, brain, have you only read the first paragraph of any popular science article on NS? I suggest you read to the end.

Maybe everyone should mind his/her own business, and give neural networks tasks which they solve perfectly!
For example: to recognize an image of gallows man or ghost?

So recognise gallowsmen and other mystical and not really working entities. Who is stopping you? But explain to the ignorant, do you recognize all of them on the charts with your eyes? I am sure they do not, so what can you teach NS? I guess you can only teach bad things.

No offence meant,
I just hate the word prediction, and it's a shame about neural networks..,
...and with them the country.
VBAG, it's a doctor's appointment, there's no help for you here.
 
grasn:

It is always nice to talk to a knowledgeable person.

Likewise!

Have you only read the first paragraph of a popular science article on NS? I recommend that you read to the end.

The fact is that in mathematics there is no such thing as a "forecast", it was invented by the authors of popular science articles not to frighten the average person with expressions like probability density or expectation, etc.

But explain to the ignorant person, do you recognize them all with your eye exactly on the charts?

Don't feel too bad about the darkness (there are spots in the sun too).

As for the presence of chart patterns - it is the only thing that can be regarded as one of the most reliable in the market. And just the figures on price and indicator charts alone can never be falsified. Read Nisson about the wise Japanese for a start.

P.S. I told you not to be offended, but you took offense (took a shit and spat). Keep in mind that those who are offended are the ones who get hurt.
 

VBAG, for a start, I read about the wise Japanese with their hangmen, sinkers and other forex nonsense that doesn't work. And to finish, I suggest you get out of your head all sorts of bad thoughts about recognising what you yourself can never confidently say.

I cannot help hating forecasts, I'm not much of a psychologist to say the least. Forecasts, predictions, fortune-tellers, .... What do you do when you place an order? What's the term you got in mind for yourself?

And I'm not offended at all, just a bad mood: the second day I can not find an error in the function of a long 1700 lines.

PS: Ok, I propose to stay friends, at least, we always have time to quarrel :o)

 
grasn:

What do you do when you place an order? What term have you got in mind for yourself?

I'm going to cut in a bit. If you trade with standard indicators such as MACD, RSI, Stoch and other things, you do not need to make predictions. If a signal appears - we open (if necessary, after confirmation). The predictions - only by non-standard ones, such as Fibo, Nerox and other.

I'm just in a bad mood: for the second day I can't find an error in the function with 1700 lines.

Let me give you some advice to improve your mood: break this function into functions not longer than 50 :) I haven't written functions longer than 50 lines, which don't fit on one screen (including comments and 1 line indents between logically different constructions) for a long time.

 
Grasn, the figures are just an example (the first thing that came to mind), while the issue of Japanese heritage is a long one and depends on subjective perception. There is no point in arguing. We can only divide all people into three categories:
- who haven't heard anything about it and are equally at ease with it;
- who have perceived this knowledge;
- Who are not yet ready to perceive this knowledge;

grasn wrote.
I cannot help hating forecasts, I am, frankly speaking, a shitty psychologist.
 I like to listen to predictions on TV about the weather, even about stock prices. (It is in the nature of man to want to be deceived).
I have simply stated my personal attitude to the word "forecast" when it is used in relation to neural networks.
I myself have a very optimistic view on the use of neural networks in dealing.

(several years ago I made at least 1000 sets in BraneMakere and NS2 in a year)

Grasn-Friends Forever, no question!
 
plan:
Mathemat:
There you go, plan, there's a classification after all, and that's encouraging. Thanks for the idea!

P.S. What's the grid, if it's not a secret? I still fiddled with Jordan/Elman.



Actually, I wrote everything myself. In C# :) It's easier to understand and implement something of your own. For example, I have a modified mesh learning algorithm. The meshes are multilayer (for example, 8 - 60 - 20 - 1), combined into committees: each one implements a different idea.
This is not really a classification. It's Valery Salov's idea which he proposed as MTP (maximum trading profits), it is used by several sophisticated systems which were written for banks. Except that conventionalneural network architecture doesn't work for it, we should focus on neural network commite with weighted control system.

2plan: Have you found the source code of your thesis project? :) I thought they are not available on the net ...
 
2rip: That's the committee I'm referring to. I haven't looked for source code anywhere. I just made a coursework on paralleling grids (during training, of course), so I have some experience in this area. :)
 
Mathemat:

I'm just in a bad mood: for the second day I can't find an error in a function that is 1700 lines long.

Let me give you some advice to improve your mood: break this function into functions no longer than 50 :) I haven't written functions longer than 50 lines that don't fit on one screen (including comments and 1 line indents between logically different constructions) for a long time now.


I don't write functions longer than 25 lines (including comments and 1 line space between logically different constructions) - :))

In general, the Achilles' heel of NS in trading, imho - curve fitting.
There are a lot of varied parameters, and you get into overoptimization at once. It is a non-trivial task to circumvent this problem.
 
plan:
2rip: That's the committee I'm referring to. I haven't looked for source code anywhere. I just made a coursework on paralleling grids (during training, of course), so I have some experience in this area. :)
Then let's go off-line. email: rip@ukr.net
:)
 

to Mathemat.

I will slightly interfere. If you trade with standard indicators such as MACD, RSI, Stoch and other stuff, you do not need to make predictions. If a signal appears - we open (if necessary, after confirmation). Forecasts - only on non-standard ones, such as Fibo, Neroc and others.

This is a certain philosophy. In my understanding, the use of standard indicators is the same prediction, no worse than the results of NS. The aggregate of signals tells us what bet we will make with DC, the price will go up or down. Such a forecast may not be very accurate and show only the direction, but anyway it is a forecast calculated on the basis of something. But in this case the trader blushes a bit and picks the floor with his sock, he says: "I did not wash it, I had nothing to do with it, I haven't predicted anything, they caused a loss. :o)

Let me give you a tip to improve your mood: break this function into functions no longer than 50 :) I haven't written functions longer than 50 lines, which don't fit on one screen (including comments and 1 line indents between logically different constructions).

Thanks for the advice, but still you can't lose the skill - I found a mistake (I used to be a good programmer). So, like in that cartoon - "I'll get better now". :о) I was avoiding numerous functions with one purpose - to reduce calculation time, associated with repeated calls of functions. Now I have obtained, that necessary parameters for 3000 series were computed by MT for 5 seconds (before it was about 30 minutes for such series, when I had a lot of these functions).

PS: By the way, 1700 rows is quite arbitrary. I inherited a sprawling style from my old programming, so the number of lines can be greatly reduced.

to VBAG

That's good. Let me remind you that I also expressed my thoughts about using NS and as I've already written I'm not an ardent opponent of their use. Good luck, friend :o)

1234567
New comment