Recognising images ( rhetorical theme ) - page 14

New comment
 
storm:


At constant lot the balance graph is the same. MO, with 0.1 lot >36.

I don't use Martin at all.

It doesn't matter what the profit is made of. Show me an EA without drawdown.

:) No comment.


I wonder why you posted that chart then? You're a) misinforming people with your chart, it's not a pattern trade and it's a fit, and b) you're not responding to correct comments. You have a "volume" line down there and it gives you away.

 
gip:


I wonder, why did you post this chart? You are a) misinforming people with your chart, it's not a pattern trade and moreover it's a fitting, b) do not react to the correct comments.

I posted the chart just to show that the pattern works, how effective it is is another question.

You've got a line down there called "Volume" which gives you away.

What does it give me away? I've written in that post (above report) lot depends on volatility and only. I've just said in a previous post what would happen to the chart if the lot is made fixed. (but nothing will, it remains the same - I checked it).

 

I will also make a contribution to this topic...

I can say with certainty that the most effective tools are the simplest tools (AO, ZIG-ZAG etc. etc. indices are debatable), and

To calibrate the NS recognizing images you need a different neural network-calibrator...

So... draw conclusions gentlemen...

Don't fill your brains with it and don't waste your money by ordering something with a "cool" NS name.

 
Debugger:

To calibrate an image-recognising NS, you need another neural network to calibrate it...

So... make a judgement, gentlemen...

Don't fill your brains with this and don't waste your money by ordering something with a "cool" NS name.

You do not need an NS to recognise patterns, all you need is to pass a series of candles through a filter describing the patterns. The NS is needed to make a decision to open an order after receiving a signal from the filter.
 
icas:
You do not need NS to recognise patterns, you just need to pass a series of candlesticks through a filter describing the patterns. The NS is needed to make a decision to open an order after receiving a signal from the filter.
But how to prepare these patterns knowing beforehand what will happen after this pattern? I.e. how and how will the filter know which patterns are "good" and which are "bad"?
 
joo:
And how do you prepare these patterns, even knowing in advance what will happen after this pattern? That is, how will the filter know which patterns are "good" and which are "bad"?

The filter does not need to know which patterns are good or bad. It gives a signal that there is a pattern, and then the NS decides what to do. In other words, the filter is another layer of NS.
 
icas:
.......... .I.e. the filter is another layer of NS.

the inconsistency turns out to be. ->

icas:

No NS is needed to recognise patterns, just run a series of candles through a filter describing patterns....

 
Here we mainly consider "zigzag methods" of pattern coding, but one can easily imagine alternative ones. For example, let's approximate a part of a price series not by a zigzag :), but by another function, at least by the same Fourier or polynomial and use a vector of amplitudes as a characteristic. Since everything has already been thought up before us, quite a few people must have followed this way.
 
icas:
You do not need NS to recognise patterns, you just need to pass a series of candles through a filter describing the patterns. The NS is needed to make a decision to open an order after receiving a signal from the filter.


The NS is the filter. And this is all useless rubbish.

 
I recall that Osowski has a description of encoding with a DPF.
Files:
os_p111-114.rar  501 kb
1...789101112131415
New comment