Machine learning in trading: theory, models, practice and algo-trading - page 2221

 
mytarmailS:

Well, it's okay without random.)

I still don't understand what his role is there

Creating tags, I think it's more normal than zigzagging or God knows what
 
Maxim Dmitrievsky:
Making marks, I think it's more normal than zigzagging or God knows what

dunno... maybe I don't understand something

 
mytarmailS:

I don't know... Maybe I don't understand something.

There's nothing to understand, it's not the most important part of the algorithm. You can mark it up your way
 
Wait till I get up to speed
 
Fast235:
Hold on, I'll be right in.
Get out and in nicely 🤣🤣 Okay, you guys are funny, I'll come back later.
 
Maxim Dmitrievsky:
Enter beautifully

If we still talk about tastes,

 
Maxim Dmitrievsky:

Anyway... I did

did not train anything yet but visually, I do not see anything like that


rsa on the signs of 5 comp

and with simulated data

 
mytarmailS:

Anyway... I've done

did not train anything yet but visually, I do not see anything like that


psa is on the signs of a 5 comp

and with simulated data

clusters are more dense, at least. Fewer emissions should be.

 
Maxim Dmitrievsky:

clusters are denser, at the very least. There should be less emissions.

We should try to practice...

If non-normalized prices are submitted to rsa, then you can see the difference

but it makes no sense to use non-normalized ones


 
mytarmailS:

I figured out how you can train a "minimum" neuron from almost any package, which is sharpened for classification or regression

The main thing is that the package gives access to the weights of the neuronka and allows you to change them.


The recipe is as follows.

1) train the neural network, it does not matter what, the main thing is to get a model with weights

2) choose an optimization method (genetics, muravi, swarming, annealing simulation, etc.)

3) write a fitness function

4) take the weights of the neuron and represent them as parameters for optimization

that's it!!! )))

You can train neuronka to make a profit, or make it create some kind of mega indicator or whatever...

It's all been thought of long ago! See R package - "automl"!