Machine learning in trading: theory, models, practice and algo-trading - page 3075

New comment
 
mytarmailS #:
If, for example, there is a trading system that works on history....
What stress tests can be devised for it to increase the probability that it will work on new data?

What if we try to answer this question in terms of the article that Maxim gave?

https://arxiv.org/pdf/2201.12692.pdf
 
СанСаныч Фоменко #:

What if you try to answer this question in terms of the article maxim gave?

https:// arxiv.org/pdf/2201.12692.pdf

Pride won't allow it. He did it all a long time ago in 15 minutes 🗿

 
Maxim Dmitrievsky #:

Pride won't allow it. I mean, he did it all a long time ago in 15 minutes 🗿

Did you use something to translate this article? I tried yandex, it's dead.

 
Aleksey Vyazmikin #:

Did you use something to translate this article? I tried Yandex, it's dead.

with your head. Ask for GPT, it's sometimes better contextualised.

 
Maxim Dmitrievsky #:

with your head. Ask for GPT, it sometimes adjusts better to the context

Oh, I'm behind - I don't have GPT :(((((

 
Aleksey Vyazmikin #:

Did you use something to translate this article? I tried Yandex, it's dead.

https://www.chatpdf.com/.

and then the translation.

    Welcome to this informative PDF on "Meta-learning for causal estimation"! In this article, Gabriel Okasa explores the results of a finite sample of meta-learners for estimating heterogeneous treatment effects. Read on to learn about the benefits of split sampling and cross-fitting to reduce systematic error and improve the effectiveness of meta-learners.

    Sample questions:
    What are meta-studies and how do they help estimate causal effects?
    How do sample splitting and cross-fitting reduce systematic overfitting error in meta-learner evaluation?
    What are practical recommendations for using meta-learning in small and large sample sizes?

    What are meta-studies and how do they help estimate causal effects?

    Meta-learning is machine learning algorithms designed to estimate causal relationships. They work by learning from a set of baseline learners, which are themselves machine learning algorithms that predict treatment effects. The meta-learners then combine the predictions of these base learners to produce an overall estimate of the causal effect. This approach can help reduce systematic error and improve the accuracy of causal effect estimates. For more information on meta-learning, see Section 3.2 on page 4.

    How do sample splitting and cross-fitting reduce systematic overfitting error in meta-study estimates?

    Sample splitting and cross-fitting are two methods used to reduce systematic overfitting error in meta-student assessment. Sample splitting involves dividing the data into two or more subsets, using one subset to teach the baseline students and another subset to estimate the causal effect. Cross-fitting involves training each baseline learner on a different subset of the data and then using each learner to predict treatment effects on the other subset of the data. Both methods help reduce overtraining error by preventing baseline learners from remembering noise in the data. For more information on sample splitting and cross-fitting, see Section 2.2 on page. 3.

    What are practical recommendations for using meta-learning in small and large sample sizes?
    This article presents practical recommendations for the use of meta-studies in empirical studies depending on specific characteristics of the data, such as treatment proportion and sample size. In small samples, full-sample estimation is preferred, while in large samples, sample splitting and cross-fitting are useful for reducing systematic error and the effectiveness of meta-learning, respectively. For more information on these recommendations, see Section 4 on page. 5.
 
Aleksey Vyazmikin #:

I'm behind - I don't have GPT :(((((

You should have asked him to write code for himself, not us ) even better than a student, no need to feed him sausage

 
Valeriy Yastremskiy #:

https://www.chatpdf.com/

and then the translation

    Welcome to this informative PDF on "meta-learning for estimating causal effects"! In this article, Gabriel Okasa explores the results of finite-sample meta-learning for estimating heterogeneous treatment effects. Read on to learn about the benefits of split sampling and cross-fitting to reduce systematic error and improve the effectiveness of meta-learning.

Thanks. I downloaded the file and it now prompts in English to ask questions. How do I teach it in Russian?

 
Maxim Dmitrievsky #:

You should have asked him to write code for himself, not us ) even better than a student, no need to feed him sausage

Good plan! I understand you need a foreign phone, but where do you get it?

 
Aleksey Vyazmikin #:

Good plan! I understand you need a foreign phone, but where do you get one?

Where do I get it?

1...306830693070307130723073307430753076307730783079308030813082...3699
New comment