Machine learning in trading: theory, models, practice and algo-trading - page 2549
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
But not knowing Matlab makes the task more difficult.
Matlab seems to have a way to generate C++ code
The case when "R" can be inserted in all posts, it's still off topic
the question originally sounded something like this: "who has ported LGBM models trained in python (this is important) to mql language?" if you haven't, you may keep silent :D
referring to the problem. Not a problem, just an inconvenience.
Matlab seems to have the ability to generate C++ code
Yes, I tried the built-in matlab generator a while ago and got a lot of errors.
As it turned out, I didn't have confidence in it. How it is now I don't know, I'll have to try it.
But I've made sure more than once - logic must be transferred by reading it.
The generator is for simple tasks.
Probably because he knows Python.
Everyone adapts to the language he knows.
I, for example, am a fan of C. And it's hard as hell to transfer everything from another language.
Especially from a language you don't speak ideally.
I have proprietary problems in Matlab, which I would like to transfer to C, aka mql.
But not mastering matlab makes the task harder.
What's the problem with pulling matlab from C (from MQL)?
the site had articles a la "working with MatLab".
there are other ways, there is SciLab, for the most part MatLab compatible, with a coherent C-API and no license chains
The porting of an implementation from one language to another, especially to a language that is not related, is not only laborious, but it also leads to a mountain of errors
what is the problem with pulling matlab from C (from MQL)?
there were articles on the site a la "working with MatLab".
there are other ways, there is SciLab, for the most part MatLab compatible, with a coherent C-API and no license chains
The transfer of implementation from one language to another, especially to a language that is not related, is not only labor-intensive, but it can also gather a mountain of errors
Yes, I somehow did not think it was possible to connect the two languages.
Perhaps because I do not like such a bundle, something to pull from somewhere.
And to implement a bundle, you need to study something again, and not the fact that it will work.
Because, as always, you won't be able to find information on API.
But thanks for the tip, I will think about it.
How can learning be improved if there is a known sample to evaluate learning (future) that will not participate in learning?
Please give your thoughts and ideas, and comments on why this "cheating" will not be effective on new data that has not yet emerged at all - or maybe it will!
How can learning be improved if there is a known sample to evaluate learning(future) that will not participate in learning?
Please give thoughts and ideas and comments as to why this "cheating" will not be effective on new data that has not yet emerged at all - or maybe it will!
If the future is known, there is no need to teach anyone anything anymore)
Isn't the standard approach to optimize algorithm meta-parameters based on results on a control sample?
If you know the future, you don't need to teach anyone anything.)
It seems to be a standard approach - to optimize the meta-parameters of the algorithm by the results on the control sample?
No, you can only train on the train sample, and let's say we have test - for control results and exam - for independent evaluation, so you can only apply the machine learning algorithm on the train sample by conditions. You can not add new predictors that describe the section exam - let them do the rest (for now).
Is there any way to analyze exam to improve the model built on train?No, you can only train on a train sample, and let's say we have a test - for controlling results and exam - for independent evaluation, so you can only apply the machine learning algorithm on a train sample by conditions. You can't add new predictors describing the exam section - let them do the rest (for now).
Is there any way to analyze exams to improve a model built on a train?Generally speaking, after training (on train) there is not one model, but a set of them, defined by meta-parameters. For example, different degree of interpolation polynomial or different regularization coefficients in lasso regression, etc. Then the best value for the metaparameter is determined (the best model from the set is taken by testing). In turn, optimization of meta-parameter on test can also be determined by some parameters (meta-parameters), for optimization of which exam can be applied. For example, in what proportions to divide the original sample into train and test.
But, most likely, I just do not understand your idea).
But I probably just didn't get your idea)
There are a lot of us)