Machine learning in trading: theory, models, practice and algo-trading - page 636
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
I don't know yet. I look at nonentropy as an additional parameter to Hearst, asymmetry, kurtosis, etc., and this parameter is the most mysterious and, how should I say it? - awesome, yes.
Another observation taken from experience. I took inputs only with entropy greater than zero and not much value. As a result, optimization is noticeably worse, in comparison with the set containing the same inputs and all the rest with negative entropy. Hence the conclusion. It is necessary to choose such inputs, whose entropy revolves around zero from both sides. That is, a set of inputs without negative entropy trains worse than a set that has both negative and positive entropy.
When there are ONLY entrances with negative entropy, such sets are also trained very well, but due to peculiarities of the optimizer and absence of quantity of such entrances the models fall short in comparison with the set with positive entrances. The main thing that this plus inputs was as small as possible...
Brothers, one small step left for us, but it will be a huge step for all mankind.....
I agree with you. Michael. And this is not a joke. I seriously think so.
I agree with you. Michael. And this is not a joke. I really think so.
So... There are two columns A and B, how to calculate the conditional probability of A from B? The Internet is full of formulas, but the examples are somehow wrong... I can not understand it :-(
Stop fucking around and behave yourselves.
One doesn't know what he's writing about, and the other one encourages him ))))
If you want to know everything about the future at a certain point in time, i.e. to predict, you must reduce the process to a Markovian one. Make it so that the nonentropy --> 0.
Are you sure that's the case? For non-Markovian process it is assumed that the price will behave identically (the same movement after the same patterns), you can take the current price pattern and the trained neuronics will tell you where the price will go next. This is very good.
But what should I do with the Markov process? How do I trade what is completely random?
Stop fucking around and behave yourselves.
One doesn't know what he's writing about, and the other is baiting him )))).
What if the man really ran to the finish line? Okay, I will not litter the thread with floods ^)))))
Are you sure that's how it works? For a non-Markovian process, it is assumed that price will behave the same way at least occasionally (the same movement after the same patterns), you can take the current price pattern and the trained neuronics will tell you where the price will go next. This is very good.
But what should I do with the Markov process? How do I trade something that is completely random?
Randomness is not random :-) The price movement always has a reason and, therefore, it's not correct to say that its movement is by chance. It is another matter that the observer may lack information about the cause and then the movement becomes random. IMHO....
So... There are two columns A and B, how to calculate the conditional probability of A from B? The Internet is full of formulas, but with examples as it is not so... I can't figure it out :-(
https://www.mql5.com/ru/articles/3264
Are you sure that's how it works? For a non-Markovian process, it is assumed that price will behave the same way at least occasionally (the same movement after the same patterns), you can take the current price pattern and the trained neuronics will tell you where the price will go next. This is very good.
But what should I do with the Markov process? How do I trade something that is completely random?
Imagine a Wiener model with drift - and it's done.