Machine learning in trading: theory, models, practice and algo-trading - page 2048

New comment
 
mytarmailS:

Okay... I'll keep my fingers crossed.

But most importantly, you understand why the network can't find the pattern I described? )

Because it's statistically insignificant, most likely.
 
Maxim Dmitrievsky:
Because it is statistically insignificant, most likely

let it be so )

 
mytarmailS:

let it be so )

If you take away everything unnecessary and leave about such squiggles... there is a correlation analysis for that, find all such patterns and look. It's a seed for the NS, it should be able to look for more complex things. Subtract those Machs from the prices and find similar correlations in increments. Seeds.
 
Maxim Dmitrievsky:
If you remove everything unnecessary and leave about such squiggles... there is a correlation analysis for that, find all such patterns and see. For the NS it's a seed, it should be able to look for more complex things. Subtract those Machs from the prices and find similar correlations in increments. Seeds.

That's what I want to do.

 
Pattern recognition can probably be handled by picture or speech models. But it will be a lot of hassle. You need to transform the data in a suitable way and then use pre-trained models. For pictures Gramian Angular Field, for speech I guess Kepster will do. Transfer lerning through fast.ai to do.
 
Maxim Dmitrievsky:
Because it is statistically insignificant, most likely

or the pattern itself works out 50/50))) that's why the network ignores it.

 
Alexander Alexeyevich:

If you have a pattern, it will work out 50/50))) so the net ignores it.

This is correct, when you teach the network, it will look for an authentic pattern, and as a rule in the TS pattern is a visual figure, and if it is hard to lock, then in most cases it does not give profit.)

 

Who's going to figure it out? )


 
Valeriy Yastremskiy:

What should I change to make the double click count in half a second instead of a second?

 
Maxim Dmitrievsky:

Who's going to figure it out? )

Watching Vorontsov? ))

1...204120422043204420452046204720482049205020512052205320542055...3591
New comment