Machine learning in trading: theory, models, practice and algo-trading - page 2337

[[Deleted]]

elibrarius:
Clean the matrix itself? Some covariance coefficients will change slightly. What will it do?

We should clean the data from the noise.

data through the matrix, will give less overfit

I dug a lot of awesome stuff on top of that, didn't have time to study it yet

[Aleksei Kuznetsov]

Maxim Dmitrievsky:

data through the matrix, will give less overfit

I've found a lot of other great stuff besides this one, but didn't have time to study it yet

I am not strong in python. But I didn't see anything in the code (sections 2.6 - 2.8), where the data itself is corrected by the denoising matrix.

[[Deleted]]

elibrarius:
I'm not good at Python. But I didn't see anything in the code (sections 2.6 - 2.8), where the data itself is corrected by the denoised matrix.

I haven't got into details yet, here is a much clearer description

https://hudson-and-thames-portfoliolab.readthedocs-hosted.com/en/latest/estimators/risk_estimators.html#de-noising-and-de-toning-covariance-correlation-matrix

SectionDe-noising and De-toning Covariance/Correlation Mat rix

This is probably more suitable for portfolio strategies

[[Deleted]]

With this R, people were stirred up, while Python is the perfect language for reserch. It's a good thing I didn't succumb

[Aleksei Kuznetsov]

Maxim Dmitrievsky:

I haven't looked into it in detail yet, but here's a clearer description

https://hudson-and-thames-portfoliolab.readthedocs-hosted.com/en/latest/estimators/risk_estimators.html#de-noising-and-de-toning-covariance-correlation-matrix

De-noising and De-toning Covariance/Correlation Matrix

This is probably more suitable for portfolio strategies

I didn't see any correction of the data here either.
I do not see correction of data here either.)

[[Deleted]]

elibrarius:

Here, too, did not see the correction of the data itself.
Apparently it seemed)

Obviously there is an inverse transformation to this. Otherwise it makes no sense.

[Mikhail Mishanin]

Applied Prado Fan Article https://dou.ua/lenta/articles/ml-vs-financial-math/

[Aleksei Kuznetsov]

Maxim Dmitrievsky:

Obviously, there is an inverse transformation to this. Otherwise there is no point.

I don't see it in the code(
Maybe they just drop the correlated (after de-noising) instruments from the portfolio... They keep talking about portfolios.

[[Deleted]]

elibrarius:
You can't see it in the code(
Maybe they're just dropping correlated (after de-noising) instruments from the portfolio... They keep talking about portfolios.

I think it's all called Agglomerative Filtering\clustering. I can't say anything without studying the subject, but it's interesting :)

the codes in his book are copies of arxiv papers

[[Deleted]]

Mikhail Mishanin:
Applied Prado fan article https://dou.ua/lenta/articles/ml-vs-financial-math/

Yes, but there is no mention of filtration