Machine learning in trading: theory, models, practice and algo-trading - page 3393

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
You don't need to put complex code in there.
.
Well, why do you need it?
1) the idea, code, preprocessing should be done in one language (R, python, JS, c++, c# ......).
2) train ONNX model in another ( python )
3) write the robot in the third language (MQL5) and then rewrite the entire preprocessing in MQL5, which is 95% practically impossible if the preprocessing is at least of medium complexity.
Anyway. total useless crap if you don't train some unworkable primitive stuff.
What's the point?
1) the idea, code, preprocessing should be done in one language (R, python, JS, c++, c# ......).
2) train ONNX model in another ( python )
3) write the robot in the third language (MQL5) and then rewrite the entire preprocessing in MQL5, which is 95% practically impossible if the preprocessing is at least of medium complexity.
Anyway. total useless crap if you don't train some unworkable primitive stuff.
What do you mean by medium complexity preprocessing? Pipeline can also be put into ONNX.
They never added onnx to R? Why are they so slow?
1) what is meant by medium complexity preprocessing? Pipeline can also be pushed to ONNX
2) R onnx has never been added? Why are they so slow?
1)
Forexample, I have my own code in some language that accepts an array of prices from MT5.
Then I build a complex TS (in my language) with thousands of lines of code, with complex logic + tracking of positions + use a dozen of different libraries. (this is what I call preprocessing).
And then only as a filter of TS signals comes some kind of AMO, that's only 3% of the code.
ONNX is only about models and only those they have implemented there.
Yes, it is possible to add your own custom code, but as far as I read it is not so easy and there is very little information on this case.
2)
Not added, I read that they do not see the demand for this technology in the p-makers, but there is a seamless bundle P + python so that in prinzepe you can make any blue, but not the point.
I am bombing not out of resentment that ONNX is not available in R, but because I don't see a way to implement my complex code through ONNX.
If you have OHLC as an input, you can implement on ONNX arobot, let's say a netter, which sees its positions, looks at its balance, makes some complex calculations and gives signals in MT5
. I understand that why do such a thing if there is MT5, but this is a question to assess the limitations on the complexity of the code in ONNX .
1)
Forexample, I have my own code in some language that accepts an array with prices from MT5.
Then I build a complex TS (in my language) with thousands of lines of code, with complex logic + tracking of positions + use a dozen of different libraries. (this is what I call preprocessing).
And then only as a filter of signals of the TS comes some AMO, that is only 3% of the code.
ONNX is only about models and only those they have implemented there.
Yes, you can add your own custom code there, but as far as I read it is not so easy and there is very little information on this case.
2)
Not added, I read that they do not see the demand for this technology in the p-makers, but there is a seamless bundle of P + python so that in prinzepe you can make any blue, but not the point.
I'm bombing not out of resentment that ONNX is not in R-key, but because I don't see a way to implement my complex code through ONNX.
If you have OHLC as an input, you will be able to implement on ONNX arobot, let's say a netter, which sees its positions, looks at its balance, makes some complex calculations and gives signals in MT5
. I understand that why do such a thing if there is MT5, but this is a question to assess the limitations on the complexity of the code in ONNX .
It's just more fun to throw models at each other.
Well, that's exactly what it's designed for, just convenient transfer, just models ... everything else is crutches.
There is a huge community of JS proggers with their own ideas and JS scripts.
To get started, they need to learn
mql5 + python + ONNX.
Instead of ONNX, they could have made Docker or something similar and would have immediately attracted a lot of new customers, but they are interested in some crap....
Instead of making it easier for people to get to themselves, they make it harder...
Whatever, it's none of my business.
There is a huge community of proggers traders on JS with their ideas and scripts on JS
They're the ones who need to learn to get into
mql5 + python + ONNX
and they could instead of ONNX make Docker or something similar and would immediately capture a lot of new callers, but they are interested in some kind of c rap....
Instead of making it easier for people to get to themselves, they make it harder...
Whatever, it's none of my business.
And where do they trade through js? Apis
There is a huge community of proggers traders on JS with their ideas and scripts on JS
They're the ones who need to learn to get into
mql5 + python + ONNX
and they could instead of ONNX make Docker or something similar and would immediately capture a lot of new callers, but they are interested in some kind of c rap....
Instead of making it easier for people to get to themselves, they make it harder...
Whatever, it's none of my business.
Looking at my code.
Several models are sitting in the middle(?) of all the R code. If I take the models out of the R code and put them somewhere else, it will be a completely different code that will have to be debugged all over again!
And why?
There is µl and R with obvious functional separation of TCs. The mcl and R bundle works stably ..... and where does ONNX fit in here ?
I'm looking at my code.
Several models are sitting in the middle(?) of all the R code. If you take the models out of the R code and put them somewhere else, it would be a completely different code that would need to be debugged all over again!
And why?
There is µl and R with obvious functional separation of TCs. The mcl and R bundle works stably ..... and where does ONNX belong here ?