Machine learning in trading: theory, models, practice and algo-trading - page 2665

New comment
 
Maxim Dmitrievsky #:

when it is possible to mark not only other tools, but also the features we are learning on, to make a kind of filter on them. You can increase the correlation of a feature with the target through this. One way.

It's all already written into the wooden models...

a rule is a filter, it just picks out the most tidbit from a feature and leaves only it, the ensemble of rules (filters) is a model.

 
mytarmailS #:

It's all sewn into the wooden models...

a rule is a filter, it just picks out the most tidbit from the feature and leaves only it, the ensemble of rules (filters) is a model.

No, there is an improvement at the expense of complexity, which leads to overfitting. And here we reduce the complexity of the model through filtering.
 
Maxim Dmitrievsky #:
No, there we are improving at the expense of complexity, which leads to overfit. And here we're reducing the complexity of the model through filtering.

You can think of complexity in many ways.

Complexity is the word complexity.

if you add 100 rules instead of one, the complexity is higher in the first case, but there is less noise, remember the sine waves from the video.

 
mytarmailS #:

you can mean complexity in many ways

complexity

If you add up 100 rules instead of one, the complexity is greater in the first case, but there is less noise, remember the sine waves from the video.

And the word add up doesn't necessarily have a mathematical meaning. You can stack weapons, like pile them up, but not add them up. Put them side by side, like stack them in a pile. Increase the complexity or complexity

 
Maxim Dmitrievsky #:
And the word add up doesn't necessarily have a mathematical meaning. You can stack weapons, like pile them up, but not add them up. Put them side by side, like stack them in a pile. Increase Complexity or Complexity

That's why I don't understand your point

 
mytarmailS #:

That's why I don't understand your point

Complexity is the analogue of complexity
 
Maxim Dmitrievsky #:
Complexity is the analogue of complexity

probably

 

It is also possible to filter using standard DSP methods

but of course there will be a delay


 
mytarmailS #:

can also be filtered using standard DSP methods

but of course there will be a delay


And if the Eurobucks is growing and other correlated pairs are falling. You can't filter this out, but logically, you should prohibit trading at that moment or make a corresponding markup. Your chart won't show that. For example, the Global is trending down, and the Eurobucks is rising in the moment, but then it will also fall. You will get additional markings that do not correlate with the trend, scattered by noise. But this is just philosophy.

 
Maxim Dmitrievsky #:

The global trend is down, and the Eurobucks is rising at the moment, but then it will also fall.

It's not a fact that it will fall, maybe at this very moment the euro was ahead of the global trend.
And what is a global trend? Where are the statistics that it even exists and works, and even if it works, where is the guarantee that at the moment when you saw that the GT is down, it will not go up, as it always happens....

Well, I'm very sceptical on this subject and I've justified my position.
1...265826592660266126622663266426652666266726682669267026712672...3700
New comment