Machine learning in trading: theory, models, practice and algo-trading - page 3326
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
The reason is simple, as intended - the signals disappear because the signals are outside the narrow acceptable range on the new data.
Well, it can be compared to classification, there are clear known patterns and there are obscure unknown ones. As time goes by, there are more and more unknowns and there is nothing left in the "known" class.
Thanks for the clarification.
That's why I asked for a quote....
At the moment our thoughts are in agreement about the result. Yes, I expect that there will be a very thin sample, but as I understand the process is iterative, which means that we can know the measure and stop much earlier and use the same data to build the same wood models, which will have fewer splits and more reliable indicators in leaves.
Do I understand correctly that the initial centres are randomly located?
#32100
#32098
#11831
Looked at the compactness profile as well. It could be that it would be even more expensive than a correlation matrix. Not only in memory, but also in computation time.
If by Saber's method, memory-efficient.
I solve such a problem very quickly, I won't tell you how, because you will start calling kolkhozniks unfounded names again.
Well, you will also face the fact that your datasets will often degenerate to zero.Are you trolling me? There's only one theorem in the text and it's about calculating CCV via profile.
I'm not trolling. I don't even remember any such instances on my part here. I just didn't realise.
You are right, indeed there it is about theoretical proof of similarity of profile calculation with average (over all splits) error on"complete cross-validation " control.
I hadn't heard that term before, now I realise it's basically all possible sampling combinations.
Ok, but how sampling partitioning packages can help here is an idea I can't understand yet.
#32100
#32098
#11831
Didn't realise what those links were referring to....
But, well, the plan is to test random sampling partitioning as a way to get used to python. Besides, it turned out that CB already implements a similar idea that I wanted to write....
I also looked at the compactness profile. It may be that it will be even more expensive than a correlation matrix. Not only in memory, but also in computation time.
If by Saber's method, then memory-efficient.
I solve such a problem very quickly, I won't tell you how, because you will start calling kolkhozniks unfounded names again.
Well, you will also face the fact that your datasets will often degenerate to zero.I'm not in a hurry, I have computing resources - even if they are old, but I have 128 gigs of RAM. I want to try different methods, also to compare my own approach.
Data scarcity is a constant problem for me, and yes, it gets worse when excluding examples.
I don't understand what these links refer to.....
But, well, I'm planning to test random sampling partitioning as a way to get used to python. Besides, it turned out that CB already implements a similar idea that I wanted to write....
I'm not in a hurry, I have computing resources - even though they are old, but I have 128 gigs of RAM. I want to try different methods, including to compare my own approach.
Data scarcity is a constant problem for me, and yes, it gets worse when excluding examples.
References to the fact that I throw in topics from time to time, then people through denial or unconsciousness start to get to it and come round to it after a while
and everyone experiences some kind of specific suffering in the process.
references to the fact that I occasionally throw in topics, then people through denial or unconsciousness start to get to it and come round to it after a while
everyone experiences some specific kind of suffering in the process.
Well that's normal in general. The same can be said for you ;)
I already, like, two years ago did a big experiment on training at different sites and selecting predictors based on the results of training, the same cross validation in essence, but without breaking the sequence of events. I heard that there is a package that allows you to save sequences so that the validation would not be affected by data before the train and the train. Do you know what it is called?
Well that's normal in general. The same can be said for you ;)
I already, like, two years ago did a big experiment on training at different sites and selection of predictors based on the results of training, the same cross validation in essence, but without violation of the sequence of events. I heard that there is a package that allows you to save sequences so that the validations would not be affected by data before the train and the train. Do you know what it is called?
I don't know what the packages are.
don't you think that forex in RF has given up for a long time, the same will happen with this resource, which is being reoriented to Asians.
so we need to learn how to apply neural networks to something else.
I don't know what kind of packages
don't you think that fora in RF has given up for a long time, the same will happen with this resource, which is being reorientated to Asians
so we need to learn how to apply neural networks to something else.
In RF there are offices with licences for FOREX - long ago I looked at the conditions - they were not very good, but if there are relatively few transactions, it may be an option.
Besides, I am more interested in the Moscow Exchange. Although, after the famous events I have not been there yet. I am interested in options, it seems that there is an opportunity to use MO there.
If I do something else, it is already work for a salary, as it is difficult to imagine that there will be a chance to earn on my labours. Even if it is something new, there is a lot of competition and the public product will be copied and the implementation will be taken apart.
In RF there are offices with licences for FOREX - long ago I looked at the conditions - they were not very good, but if relatively few transactions, it may be an option.
Besides, I am more interested in the Moscow Exchange. Although, after the famous events I have not been there yet. I am interested in options, it seems that there is an opportunity to use MO there.
If I do something else, it is already work for a salary, as it is difficult to imagine that there will be a chance to earn on my labours. Even if it is something new, there is a lot of competition and the public product will be copied and the implementation will be taken apart.
That sounds so hopeless.
Come visit us in the far east, just now I was on the embankment, the china across the river, I should have taken a picture and posted it here.
Yesterday I was also thinking of going to Harbin for a week, the price starts at 39 grand.)
Fucking country, you could drive all of Europe for that money.
A week ago, we drove 1500km in 20 hours together.