Machine learning in trading: theory, models, practice and algo-trading - page 763
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
Is it possible to do as I described above? I know that there are committees from networks, but I've never worked with them. And I assume that training should be carried out in parallel... Or it is still possible to separately? What do you think?
Yes this is the case, it is a committee, but enough ogibike in the training with one grid and the end result is not going to wait (worse). The more elements in the system the more unreliable it is. the system should be simpler.... But in general this approach may well be. It doesn't break any serious laws in networking, it only increases the probability of error due to a large number of elements.
5 nets need to be trained so that all 5 of them are trained correctly. If one of them starts to fidget, it may affect the deterioration of the result as a whole. You can do it, but I would try to reduce number of inputs...
If the input data is a lot of 1000 for one answer and not enough 1000 such examples, then most likely the significance of the variables in the example is very weak and noisy. the number of examples I personally use from 10 000 , and with such a number the random forrest algorithm is quite good. and answering your question about the results of 5 grids then send to the 6th grid yes, you can train that way. But even here you can take a slightly different approach, just put the importance of these 5 grids on the ballot. that is, as a total prediction to use. and once again, think carefully about the input data.
thanks for the explanation... I will think...
Michael, do not destroy the soul of the Orthodox... Make the Grail faster. I can`t take it anymore...
About working with time series from a financial dataset scientist
It may be useful to someone. For those who have a Kaggle account
Numerai is giving away $1 million in Numeraire to Kaggle users.
Visit numer.ai/airdrops to claim 30 NMR (~$500) by linking your Kaggle account.
I got the same message. The nuance is that these NMR tokens will pay out in chunks for months, and the price of NMR is dropping.
And the Kaggle account has to be higher than a beginner.
p.s. There is some kind of mess going on there in the numeraire. Then the results sent a few hours before the end of the contest refused to be accepted and remained unprocessed. Or a model based on a randomly drawn set of predictors was suddenly "unoriginal", even though the chance for unoriginality was almost zero. It was also unclear about the stakes, and sometimes all of them were losing.
The feeling that twice a year they have a demonstrative update, everyone ran to buy their tokens and lost them in bets. And it goes around and around.
Michael, do not destroy the soul of the Orthodox... Make the Grail faster. I can't take it anymore...
In fact, the Grail is different for everyone, I found mine. Now I'm looking for investors, because I don't want to spend money here and there. I will organize a PAMM account, so you're welcome. I have no need to do anything at all. I need to have the money :-)
I have a lot of money in my account. And that's how you make money from the trades, and then bam!!!! And the second shift :-(. Two trades put my account overdrawn only because one motherfucker, whom I see in the mirror in the morning, decided to overdraw stops and did not train the AI on the new futures.... It's still the same model trained when the fuck knows when
Semyon Semyonych.....
In fact, the Grail is different for everyone; I have found mine.
I agree. A great saying.
I suggest a little prayer and patching up the money bags while my grail is being tested (very slowly without the GPU)
:))
By the way the tests.... let's get into the fray.... see what happens....