Machine learning in trading: theory, models, practice and algo-trading - page 1019

[mytarmailS]

Vladimir Perervenko:

And we don't do anything else but time series. No?

And there are a lot of methods, and not only of choice, but of creation as well. There is a whole new section of Features engineering. Look at several articles where it was discussed in detail.

Good luck

Vladimir, I didn't mean classification, I meant working directly with BP. For example (just an example from the ballpark)

Algorithm breaks BP into components, say for 20 (1,2,3....20) and starts to try all possible combinations between them, in the course of which it turned out that if the series leave only the 1st, 5th and 13th that is, all three components, the error in the new data is less than 40%.

Once again I say it's just an example, and it's not a classification, are there any algorithms "future section" but for BP directly?

[Aleksey Terentev]

Bookmark it.

[Vladimir Perervenko]

mytarmailS:

Vladimir I did not mean classification, but to work directly with BP. For example (just an example from the ballpark)

The algorithm divides BP into components, let's say into 20 (1,2,3....20) and starts to try all possible combinations between them, in the course of which it turned out that if in a number to leave only the 1st, 5th and 13th that is all three components then the error in the new data is less than 40%.

Once again, I say this is just an example, and it's not a classification, are there any algorithms "future section" but for BP directly?

I have no idea what you are talking about. A time series of what do you want to break down into components?

[Vladimir Perervenko]

Aleksey Terentev:
Bookmark it.

For TensorFlow/keras, the information is very outdated. But in general useful info.

[Vladimir Perervenko]

Python's development system is not impressive at all. Is there a future?

[mytarmailS]

Vladimir Perervenko:

I have no idea what you're talking about. A time series of what do you want to break down into components?

OK, forget it.

[Vasily Perepelkin]

Zhenya:

I guess you're right, look here in 2-3 years, when the new generation will come, who knows, maybe they will be smarter, though ... I highly doubt it. Astalavista losers! Write articles and riveting $50 graals.

Goodbye, we do not need your tester and a $50 grail, I agree with my student Maxim Denisenko

[Ivan Negreshniy]

Vasily Perepelkin:

I agree with my follower Maksim Denisenko on this point.

In fact he suggested an excellent idea to unify formats and practical exchange of models, maybe this is the only way to revive the MO theme for trading, since theorists seem to have already lead it to a deadlock.

It's just a pity that I didn't manage to develop this idea - I started as a great combinator and got off like a pussy...

[Vasily Perepelkin]

Ivan Negreshniy:

In fact, he expressed a great idea to unify formats and practical exchange of models, perhaps this is the only way to revive the topic of MO for trading, since it seems that theorists have already brought it to a deadlock.

It's a pity he failed to develop this idea, he started as a great combinator and ran away as a pussy...

For example there is no connection between the trajectory of a ball on a soccer field 10 minutes ago and the current one, or between the way you are driving a car, the patterns of accelerations and reversals, they are not predictors of future accelerations, U-turns and trajectories, the dynamics of the price of its "pattern" is only a unique and unique past, it does not affect the future, it's all like a fucking omens and zabubon about a black cat, dressed in mathematics, quit it before it's too late.

[Ivan Negreshniy]

Vasily Perepelkin:

This whole topic is a dead end, the market is not ruled by numbers, it is ruled by people, human psychology, for example there is no connection between the trajectory of the ball on a soccer field 10 minutes ago and what will happen now, or how you drive a car, the patterns of acceleration and reversals, They are not predictors of future accelerations, U-turns and trajectory, the dynamics of the price of its "pattern", this is only a unique and unique past, it does not affect the future, it's all like a fucking omens and black cat bells, dressed in mathematics, quit it before it's too late.

Well, if there were no numbers there would be no market, at least in its current form and digitized ball trajectory is quite related, even people sooner or later will digitize and will teach them not by howling, like you, but by uploading information directly into their neurons;)))