Machine learning in trading: theory, models, practice and algo-trading - page 2391

[iwelimorn]

Maxim Dmitrievsky:
by the way, python 3.9 is significantly faster than 3.8 in the console, switched to it

I have no such fierce tasks as counting tons of profits with high frequency, so for now 3.7))

[iwelimorn]

Maxim Dmitrievsky:

it is the second model that allows/denies to open trades

i.e. there are two models in the production.

I see. I have a discriminator. Some time ago I successfully implemented a script that collects the best models on various attributes and parses them into one EA.

Maybe a meta-model should be applied to several optimal generative models at once?

I'll have to try it

[Maxim Dmitrievsky]

I see. I have a discriminator. Some time ago I've realized a script that collects the best models on different attributes and parses them into one Expert Advisor.

Maybe the meta-model should be applied to several optimal generative models at once?

It is also a discriminator, i.e. a bunch of 2 or more models is simply retrained.

But I have not implemented the loop yet, all f-and-s have to be redesigned there.

maybe even a few, I don't know yet.

[iwelimorn]

Maxim Dmitrievsky:

it is also a discriminator, i.e. it simply retrains a bunch of 2 models

But I haven't implemented the loop yet, I have to redo all the functions there.

I can send you my scripts, maybe help what.

[Maxim Dmitrievsky]

I can send my scripts, maybe help what.

It seems to be clear how to do

[Evgeni Gavrilovi]

Maxim Dmitrievsky

I can write an article

That would be great) I'm always happy to see articles from you.

[Maxim Dmitrievsky]

Evgeni Gavrilovi:

That would be wonderful) I am always glad to see articles from you.

I always enjoy something new myself, but it gets harder and harder to come up with every time )

As long as there are no qualitative changes compared to the previous one, i don't see the point of writing

[Evgeni Gavrilovi]

Maxim Dmitrievsky:

I always enjoy something new myself, but it gets harder and harder to come up with every time )

As long as there are no qualitative changes compared to the previous one, I don't see the point of writing

Can you share an example how to use this deep neural network instead of GMM?

https://www.tensorflow.org/api_docs/python/tf/estimator/DNNClassifier

it should be replaced

gmm = mixture.GaussianMixture(n_components=75, covariance_type='full').fit(X)

to this: tf.estimator.DNNClassifier

[Maxim Dmitrievsky]

Evgeni Gavrilovi:

Can you share an example of how to use this deep neural network instead of GMM?

https://www.tensorflow.org/api_docs/python/tf/estimator/DNNClassifier

you need to replace

to this: tf.estimator.DNNClassifier

no way, they are absolutely different things

[Evgeni Gavrilovi]

Maxim Dmitrievsky

Is MLPClassifier not suitable for this problem too?

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