Machine learning in trading: theory, models, practice and algo-trading - page 3096
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Have had a similar one for over a year, haven't merged yet.
Read (already gave a link before), will be at least someone to chat with. The handbook is mega cool, I read it. Especially section 22. Explanation of Chernozhukov's approach.
Sanych's been talking unreasonable bullshit.The video didn't impress me. It is not quite clear how advertising influence on customers (treatment), which changes their behaviour and, for example, the income of the shop --- relates to trading. It's not like we can somehow influence the participants of trading on the market (unless you have a billion to shake currencies and from that evaluate the impact of your treatment).
With our money we can only influence our balance. No impact on the part of the visitors (which will be compared later) - we can only have no trading, i.e. no change in balance.
These are my first thoughts after watching the video. Because of them I was not interested.
But your balance is interesting. It is possible that you do not apply the whole method without changes, but take separate elements.
So, I'll still read it (maybe I can do it in a week, in between working moments).
The backstory started with my last article"Model Meta...", where I decided to use model errors to adjust the model. Then rewrote it in various other ways. Then I found out that kozul inference is about the same thing and there are no other adequate ways to improve models.
Model errors are corrected by any bousting, on every new tree after the first one.
Maybe it's all about improving error correction, not about kozul inference?
The video did not impress me. It is not quite clear how advertising influence on customers (treatment), which changes their behaviour and, for example, the income of the shop --- relates to trading. It's not like we can somehow influence the participants of trading on the market (unless you have a billion to shake currencies and from that evaluate the impact of your treatment).
With our money we can only influence our balance. No impact on the portion of visitors (which will be compared later) - we can only have no trading, i.e. no balance sheet change.
These are my first thoughts after watching the video. Because of them I was not interested.
But your balance is interesting. It is possible that you do not apply the whole method without changes, but take separate elements.
So, I'll still read it (maybe I can do it in a week, in between working moments).
The same concept of tritment is used in econometrics for causal inference. An effect is defined as any isolated variable. You can take any attribute and declare it a tritment and see how it affects forecasts. One variable will be a tritent, the rest will be nuisance parameters. Such definitions are not difficult to memorise, I think.
Model errors are corrected by any booleaning, on every new tree after the first one.
Maybe it's all about refining your error correction, rather than cosool inference?
Booosting is overfitting for noise. And here the opposite goal is for there to be no overfit. If there is no overfit, there is cause and effect, in simple terms.
Bousting is overfitting for noise. And here the reverse problem is that there should be no overfitting. If there is no overfitting, there is cause and effect, in simple terms.
The task is good. I will read, and if there is something interesting and unclear, I will ask.
The topic is not simple, it is not the usual concepts of MO. There is a junction of statistics and MO. I haven't learnt everything myself yet.
It won't fly from the start.
Have had a similar one for over a year, haven't merged yet.
Read (already gave a link before), will be at least someone to chat with. The handbook is mega cool, I read it. Especially section 22. Explanation of Chernozhukov's approach.
It doesn't use any "packages", everything is written by yourself. So don't be intimidated by python.
Because Sanych is talking unreasonable rubbish.
The backstory began with my last article"Meta model...", in which I decided to use model errors to correct it. Then I rewrote it in various other ways. Then I found out that kozul inference is about the same thing and there are no other adequate ways to improve models yet.
Some childish offences prevent to understand the sense of what I said, which (sense) is banal: everything that is presented in the article as a discovery I personally did 5-6 years ago, I was not the discoverer at that time, before me was chewed and chewed. Really didn't know it was "cajual" and didn't introduce new terminology on trivial things.
Let's drop it.
Can you show here at the code level the meaning of"kozool inference" with predictors, target, model, estimation and other "newness" or whatever? This will be a proof of the novelty of the "kozul inference" direction,and the article is not such a proof, though I liked its results - I myself killed a lot of time on obtaining similar results with absolutely stupid work.
Some childish offence prevents me from understanding the meaning of what I said, which (the meaning) is banal: everything that is presented in the article as a discovery I personally did 5-6 years ago, I was not the discoverer at that time, it had been chewed and chewed before me. I really did not know that it was "cajual" and did not introduce new terminology on trivial things.
Forget it.
Can you show here at the code level the meaning of"kozool inference" with predictors, target, model, estimation and other "newness" or whatever? This will be a proof of the novelty of the "kozul inference" direction,and the article is not such a proof, though I liked its results - I myself killed a lot of time on obtaining similar results with absolutely stupid work.
You called it a branch of statistics (not me) - kozul inferens. You either accept it, or you will continue to torment yourself and others with meaningless questions.
You opened something you don't understand? :) and now you are asking for explanations. I don't run training courses. Your comment on the article was rambling, there is nothing more to discuss here so far. I have not given enough sources for self-study? Should I recount dozens of pages maybe in 2 words, or better yet throw a bag at you? I don't know how you're used to it.
You called it a branch of statistics (not me) - kozul inferens. You either accept it, or you will continue to torment yourself and others with meaningless questions.
Thanks, got it, specific example will not be, "take blank"
Thanks, got it, the specific example won't be: "take blank"
You're a big fan of Microsoft, aren't you?
https://github.com/py-why/dowhy
Here, read the case studies. For some reason kolkhozniks from Microsoft made a no-one needs a library on kozulu. I'm shocked myself, Sanych said that it's a kolkhoz, probably they didn't get it to them.
I don't understand what a man needs. Do you need to chew it up and put it in your mouth?