Machine learning in trading: theory, models, practice and algo-trading - page 676
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
No, Maxim - I think you only used one type of input data - increments on the minutes. Right? And you need to try different ones!
And work exclusively with one software product, which is very important.
Then if you are not too lazy, summarize the results of experiments in one table with the specific data of the model and the software product and publish it again - for memory and reflection.Different indicators, oscillators, other currency pairs, decomposition of quotes by frequency, increments, AR models, VAR
I have fed the equity chart and the results of previous trades
Well, it's clear that this is an amateur level and you can't try everythingVarious indicators, oscillators, other currency pairs, decomposition of quotes by frequency, increments, AR models, VAR
I fed the equity graph and the results of previous deals
All this in a specific program? Which one?
You see, the thread got too vague or something, I don't know how to say it. You need kind of a bottom line - a general summary table of prediction results. This I have not seen. Therefore I consider a branch incomplete, but has lost its original meaning.
And this is all in one particular program? Which one?
In MQL5 :)
alglib package is built-in, there is an NS
library
in MQL5 :)
alglib package is built-in, there is the NS
library, or rather
Can you definitely say, for example, that this package has the best/worst predictions with such an input data?
So that people, in the future, won't touch it? To sum it up?
You see, the branch is too blurred or something, I do not know how to say it. You need a kind of bottom line - a general summary table of the prediction results. This I have not seen. Therefore, I consider the branch incomplete, but has lost its original meaning.
So start a new thread, and if the topic of interest, people will pull up. With the theory and practice already have experience).
Can you definitely say, for example, that this package made the best/worst predictions with this input?
So that people, in the future, won't touch it? Well, to sum it up, as it were?
I have nothing bad to say about the package, everything works, no bugs detected
I have no complaints about the libraries themselves, but I have complaints about myself or the approach in general
so let them be drawn, but with their own ideas :)
the main problem of all NS stated here is overlearning
the main problem of all NS stated here is overtraining
Rather, that the result of learning is close to 50/50
The main problem of all NSs here is overtraining
What was the maximum number of neurons used in your NS? What were the structures of the NS?
In fact, as far as I understand it, the topic can be considered exhausted.
Long live the distributions and their tails!!!!! :)))))))))))))))
It's more likely that the learning outcome is close to 50/50
Well, that's when it's really bad ))))