Machine learning in trading: theory, models, practice and algo-trading - page 3237

[Maxim Dmitrievsky]

You can do in the same R, save the model, and through python in onnx in a few lines

Or through c++ I guess there's one too.

I'm interested to see how many real moshnikovs there are and what the results are.

[Andrey Dik]

Aleksey Nikolayev #:
But my namesake, apparently, intends to cram some of his own craft into the model) Logically, for 15 thousand USD is not on the road).

it seems that to avoid problems with converting homemade meshes to ONNX, it is necessary to use ONNX.... right away, already at the stage of training.

[Andrey Dik]

Aleksey Nikolayev #:
Yes and they will accept models, not exechnics like, and then they will run the models in the same template for all.

If that's the case, it's sad. not only the signalling component is important, but also the trading component based on the signals.

Some models are trained to work with stops, and some are reversal models with opposite signals....

i.e. one and the same model can be traded differently, both in profit and loss.

[Aleksey Nikolayev]

Maxim Dmitrievsky #:
You can do in the same R, save the model, and through python in onnx in a few lines

Or through c++ I guess there is also

I am interested to see how many real moshnikovs and what the results are

If only the final model itself, I guess it could work. But I'd like to learn how to stuff all intermediate transformations from price to signs into the model as well. Apparently, this is possible only in Python.

In any case, there will be at least some popularisation of the topic, which may lead to the appearance of fresh blood in the branch (and to the growth of the volume of madness in it).

[Aleksey Nikolayev]

Andrey Dik #:

It seems that in order to avoid problems with converting homemade meshes to ONNX, it is necessary to use ONNX.... right away, already at the stage of training.

About grids sounds ambiguous) Neuro-grids or order grids).

[Andrey Dik]

Aleksey Nikolayev #:
About grids sounds ambiguous) Neuro-grids or order grids?)

neuro, of course.)

[Aleksey Nikolayev]

Andrey Dik #:

if that's the case, it's sad. because not only the signal component is important, but also trading on the signals.

Some models are trained to work with stops, and some are reversal models for opposite signals....

i.e. one and the same model can be traded differently, both in profit and loss.

Well, it is obvious that the main goal is to advertise the platform and the introduction of onnx is a great reason for that.

[Aleksey Vyazmikin]

Andrey Dik #:

If so, that's sad. because not only the signal component is important, but also trading on the signals.

Some models are trained to work with stops, and some are reversal models for opposite signals....

i.e. one and the same model can be traded differently, both in profit and loss.

There will be a template MQL5 code of the Expert Advisor, which contains all the rules. I understand that the model will have to give buy and sell signals, and apparently, we also need to wait, or at least the threshold of activation of the trading signal that would be in the template.....

[Andrey Dik]

Aleksey Nikolayev #:
Well, obviously the main purpose is to advertise the platform and the introduction of onnx is a great excuse for that.

Aleksey Vyazmikin #:

There will be a template with all the rules in it. I understand that the network will have to give buy and sell signals, and apparently, you need to wait, or at least the threshold of activation of the trading signal that would be in the template.....

yes, but how will the participants' models function in the template of the organisers? will all models trade on one bar? - At the opening of a bar open a position and at the next opening of a bar close a position?

in short, I have more questions than I understand. until the organisers' template appears, it is impossible to plan anything, in my opinion.

[mytarmailS]

Maxim Dmitrievsky #:
You can do in the same R, save the model, and through python in onnx in a few lines

Or through c++ I guess there is also

I'm interested to see how many real moshnikovs and what the results are

This is if you mean only neural network model and not any model like Forest.

Although hgbust is probably possible too.

---------------------------------

And it says that you can't convert any model, the model itself must support this format.

--------------------------------

So the conclusion is that ONNH is python, no way out.