Machine learning in trading: theory, models, practice and algo-trading - page 3531
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
So the ones that were clustered, they're not in the sample? I don't understand the oddity of the results myself.....
I mean, you can. It's just that these are two different approaches, not excluding each other, but in my concept adding...
Yes, that's probably how it is in English. The main thing is that you understood what I was talking about and I didn't forget it after a while.
Yes, it's not in the sample because it doesn't participate in the training of the first model.
It participates in the second model to determine the current state (cluster) of the market.
I think you and looking at your finger will have a similar reaction.....
Again your ideas about splits/quanta are not true.
It is better to write "I again did not understand why this way and not that way, I have a different opinion" - allow that you did not understand the interlocutor or even wrong - there is always such a possibility.
at P2 will remove the true numbers of these lines and they will not go in a row, but chaotically - and will not be able to form a quantum.
The simplified example deliberately omits the exact distribution of the deleted strings over quantum segments, but speaks of the consequences, which is measured through an estimate of the change in the bias of the probability of a quantum segment belonging to a particular class. Thus, even removing a few strings from the neighbourhood of a quantum segment can result in a probability bias sufficient to consider such a segment a valuable candidate for a split. Think in terms of target values rather than row numbers - there was"011010110"(green)+"01010" (part of the red range) became "01111"(green) and "00100"(blue) -"01101011001010" greyed out the conventionally departed rows, allowing a new quantum table to be created and the boundaries of the green quantum cutoff to be repartitioned to a smaller range.
You should make printouts by rows and check your ideas, otherwise you are wasting your time for nothing.... It is obvious that the order of lines will be different. But apparently it is obvious only when you look through a lot of printouts.
I don't know how you can think that a person who deals constantly and deeply with a specific topic and a specific issue doesn't understand what he is doing.... borders on insulting.
Yeah, and Alexei Nikolaev does crossword puzzles in the tram, that's understandable...
Why do I have to write in English?
I did not declare that I know it well - and Russian badly know, so what?
In general, this is the level of schoolboys, even if a person is wrong, then laugh at him.
I didn't even write in English, I wrote in transcription - I often write that way - a habit from DOS.
Thus, even removing a few lines from the neighbourhood of a quantum segment can result in a probability shift sufficient to consider such a segment a valuable candidate for a split. Think not in terms of row numbers, but in terms of target values - there was"011010110"(green)+"01010" (part of red range) became "01111"(green) and "00100"(blue) -"01101011001010" greyed out the conditionally gone rows, which allowed to create a new quantum table and repartition the boundaries of the green quantum segment towards the decreasing range.
These rows will be removed not from one quantum (and next to it) for the P2 predictor, but chaotically from many rows across the predictor range, as I have drawn. As a result, all the quanta can be prerandomised, not just one as you drew. The picture can change a lot at all.
If that's what you meant, you didn't write and draw clearly enough to be understood the first time.
These strings will be removed not from one quantum (and next to it) for predictor P2, but chaotically from many strings throughout the range of the predictor, as I drew. As a result, all quanta can prerandomise, not just one as you drew.
I think it's clearly written:
"
The quantum segment that was used to do the split becomes grey in the drawing. The quantum segment that contains the same responses (line indices) as the grey one is crossed out with a grey line. The loss of examples in the undefined region (red) is not reflected in the figure.
"
The picture can change a lot at all.
The example is simplistic, and it is clear that in practice, different quantum cutoffs of strings - candidates for deletion - fall into different quantum cutoffs, but they carry different "damage" to the probability bias and this is the criterion by which the evaluation is made. In essence those that change the EXACT picture a lot are excluded from the split candidates.
If that's what you meant, you didn't write and draw clearly enough to be understood the first time.
I think you were looking for flaws rather than merits, hence the eagerness to understand..... but I'm glad I was able to explain it to you in the end.
The example is simplified, and it is clear that in practice there are different quantum split strings - candidates for deletion, but they carry different "damage" to the probability bias and this is the criterion by which the evaluation is made. In essence, those that change the EXACT picture a lot are excluded from the split candidates.
Simplification does not always lead to better understanding. If removing one quantum on P1 was not shown as changing only one quantum for P2, but for 2-3 quanta, there would be no questions. And I probably wouldn't have written anything at all, since everything is ok.
Simplification does not always lead to better understanding.
Agreed.
If removing one quantum on P1 wasn't shown as a change of only one quantum for P2, but for 2-3 quanta, there would be no questions. And I probably wouldn't have posted anything at all
When removed at P1, the changes were at two significant quantum segments - one each for P0 and P2.
If I publish where, I will remember the misunderstanding, and perhaps add information to the illustration in the form of more quantum segments for each predictor.
#3289
You can cry to the requiem on the link :)
I must have got the date a bit mixed up, I must have chosen the wrong scale
#75
I like to remember the long term predictions