Machine learning in trading: theory, models, practice and algo-trading - page 1993

 
Maxim Dmitrievsky:

?

drunk probably... On the plus side, absolute is relatively greater than one, less than absolute is relatively less than one. The logarithm of absolute differences leads to relative differences. They just don't teach unit circle in schools nowadays. My wife sometimes works on this subject.... From there I know how difficult it is to teach the unreasonable to the unreasonable.....

 
Valeriy Yastremskiy:

drunk apparently... On the plus side, absolute is relatively greater than one, less than absolute is relatively less than one. The logarithm of absolute differences leads to relative differences. They just don't teach unit circle in schools nowadays. My wife sometimes works on this subject.... from there I know how hard it is to bring reason to the unreached.....

I wrote a simple thing to the man, without squares or anything else.

don't write too much.

 
Maxim Dmitrievsky:

I just wrote to the man in simple terms) without squares and stuff.

Don't write a word too much.

It's hard to understand people you don't know straight away. Sometimes a couple of letters are not enough to understand what the proger wants to say, and with the coder even harder. Well... I'm going... to the remote....... sorry... if anything .....

 
Valeriy Yastremskiy:

Yes, it is difficult to understand strangers in half a word. Sometimes a couple of letters are not enough to understand what the proger wants to say, and with the coder even more difficult. Well... I'm going... to the remote....... sorry... if anything .....

I wrote that it is difficult to imagine a model that would at least show inverse dependence on a tray, i.e. learning in reverse
 
Maxim Dmitrievsky:
I wrote that it's hard to imagine a model that would at least show inverse dependence, i.e. learning in reverse, on a trace

I do not understand then about the inverse dependence. if it is a constant dependence on past data, then what is complicated here. what is the inverse dependence then.

 
Valeriy Yastremskiy:

I do not understand then about the inverse dependence. if it is a constant dependence on past data, then what is complicated here. what is the inverse dependence then.

Correlation of Source and Predicted Series
 
Maxim Dmitrievsky:
Correlation of initial and predicted series

Then the model works the other way around. I realized that the question was about the increments. An inverted model, or multiplied by minus one without moduli. Although of course in the NS layers it may not be so straightforward.

 
Valeriy Yastremskiy:

Then the model works the other way around. I understood that the question was about increments. Inverted model, or multiplied by minus one without considering modules. Although of course in the NS layers everything may not be so unambiguous.

Just this metric can return negative values, yes. But in practice it almost never happens. Then you can take the value as a relative value, what's the problem. We are not great mathematicians here.
 
Maxim Dmitrievsky:
It's just that this metric can return negative values, yes. But this almost never happens in practice. Then you can take the value as a relative value, what's the problem. We're not great mathematicians here.

Negative correlation between the raw and predicted data is the wrong model. Of course it's a rare case. And the correlation between them of course is not absolute, and rather not even relative, there on the layers depends on what order diff, acceleration of what order type.

 
Valeriy Yastremskiy:

Negative correlation between the original and predicted data is the wrong model. Of course, this is a rare case. And the correlation between them of course is not absolute, and rather not even relative, it depends on the layers of what order diff, acceleration of what order type.

There are no layers there, it's a tree boosting