Machine learning in trading: theory, models, practice and algo-trading - page 2026

New comment
 
Maxim Dmitrievsky:
In a sequence, each subsequent element must be related to the previous ones, such as words in a sentence. Otherwise it's just unstructured garbage, what to look for

Fuck, you don't know what I mean... ((

 
mytarmailS:

You have no idea what I'm talking about... ((

binary search, then you can share. You don't understand what a time series with memory, a non-Markovian process
 
Maxim Dmitrievsky:
binary search, then share

calculate how many combinations it takes to search all sequences of size 10 with a range of values between 1 and 5k

))))))) try


a sequence of size 10 with a range of values from 1 to 20

That's184756 combinations.

and here's 5k !!!!!!!

 
mytarmailS:

Well, calculate how many combinations it takes to search all sequences of size 10 with a range of values inside from 1 to 5k

))))))) try it

In the loop you find one, in the rest of the array you look for a two....acct, one pass through the array
 
Maxim Dmitrievsky:
In the loop you find a one, in the rest of the array you look for a two... ah, one pass through the array

MlLLaLLaLLaLLaLLaLLaLLaLLaLLaLLaLLaLLaL .....

That's if you know to look for the one!!! And you know ?????

In real data you know nothing !!! no sequence size, no range, no proper clusters...

Maybe you need to get some sleep ? ))

 
mytarmailS:

MLEARLY.....

That's if you know the unit to look for!!! And you know ?????

In real data you know nothing !!! no sequence size, no range, no proper clusters...

Maybe you need to get some sleep ? ))

You have a generalization task, and you randomly throw in numbers. At least put them in the same positions on each line. Sleep is a sacred thing, that's what I do.
 
Maxim Dmitrievsky:
You have a generalization problem, and you randomly throw in numbers. Put them at least on the same positions in each line

I did that absolutely deliberately...

The market is not stationary and when you create a dataset by sliding window, roughly speaking all the chips will never hit their row, think about it, it's obvious

And the generalization algorithm will just look for them in the same position where it saw them on the tray...


I only left the correct sequence... It's the only objective


That's why the RNN is better, due to its memory it just slightly compensates for this floating of chips by positions in strings

 
mytarmailS:

I did it absolutely consciously...

The market is not stationary and when you create a dataset by sliding window, roughly speaking all the chips will never hit your line, think about it, it's obvious

And the generalization algorithm will just look for them in the same position where it saw them on the tray...


I only left the correct sequence... It's the only objective

Right, but if there is some regularity in the sequence, then the grid will highlight the context. And if there is randomness like you have, it won't. 4 values out of 200 doesn't add any regularity, it's part of the noise in my opinion. Well, put it in the rnn, check it, you have it in R. I'm asleep already, from my phone
 
Maxim Dmitrievsky:
True, but if there is some regularity in the sequence, then the grid will highlight the context. And if there is randomness like in your case, it won't. 4 values out of 200 doesn't add any regularity, it's part of the noise in my opinion. Well put it in the rnn, check it, you have it in R. I'm asleep already, from my phone

Okay, sleepy,

I'm drunk on coffee, I can't sleep.)

 
and the guys are only trading on the plus side
1...201920202021202220232024202520262027202820292030203120322033...3512
New comment
Reason: