Machine learning in trading: theory, models, practice and algo-trading - page 2392

[Maxim Dmitrievsky]

Evgeni Gavrilovi:

Is MLPClassifier also not suitable for this problem?

There is a method for estimating the probability of which class the sample belongs to.

https://scikit-learn.org/stable/modules/generated/sklearn.neural_network.MLPClassifier.html

These are different models. GMM is used for estimating probability density and sampling samples, and classifier classifies

Apparently you want to replace CatBoost with a neural network. But it doesn't make much sense.

[Evgeni Gavrilovi]

Maxim Dmitrievsky:

These are different models. GMM is used to estimate probability density and sampling examples, while classifier classifies

Apparently you want to replace CatBoost with a neural network. But that doesn't make much sense.

there you write that a neural network is better than GMM

https://www.mql5.com/ru/forum/356331#comment_19373237

[Maxim Dmitrievsky]

Evgeni Gavrilovi:

there you write that the neural network is better suited than GMM

https://www.mql5.com/ru/forum/356331#comment_19373237

You were talking about generative networks and autoencoders. I have tested classical versions, they are worse. I've already written in this thread before and the code was posted on the pitcher I think.

[Maxim Dmitrievsky]

Evgeni Gavrilovi:

there you write that the neural network is better suited than GMM

https://www.mql5.com/ru/forum/356331#comment_19373237

Have a look at this modelhttps://sdv.dev/SDV/user_guides/timeseries/par.html

I haven't tried it myself, I need to generate and visualize it

as far as i understand, the model is in active development, you can communicate with the developers directly

+ I have sent a new article for testing, with new ideas

[Evgeni Gavrilovi]

Maxim Dmitrievsky

Thank you.

[Maxim Dmitrievsky]

Evgeni Gavrilovi:

Thank you.

I am not installing the bible, a lot of errors. Probably not the actual version.

[Maxim Dmitrievsky]

Pulls some version of nampai, which is not installed on either the computer or google colab

Reinstalling nampai in colab worked

Some kind of crooked monsters write these libraries.

[Evgeni Gavrilovi]

Maxim Dmitrievsky:
Pulls some version of Nampai, which is not installed on a computer or google colab

This version 0.5.0 is fine.

https://pypi.org/project/sdv/0.5.0

[Maxim Dmitrievsky]

Evgeni Gavrilovi:

This version 0.5.0 is fine.

https://pypi.org/project/sdv/0.5.0

In kolab ran the last one. Need to smoke the functionality of the model on the git, you can just copy the python module. Otherwise it is not clear how it works, there is no description in the manual.

And you can't google anything about it.

[Evgeni Gavrilovi]

Maxim Dmitrievsky:

in the colab ran the last one. It is necessary to smoke the functionality of the model on the git, you can just copy the python module. Otherwise it is unclear how it works, there is no description in the manual.

from sdv.timeseries import PAR

pr_c = pr.copy()
X = pr_c[pr_c.columns[1:]]
sdv = PAR.fit(X)

I got to the fit, then the error: fit() missing 1 required positional argument: 'timeseries_data'

i think i need another format for feeding time series

https://sdv.dev/SDV/user_guides/timeseries/par.html