Machine learning in trading: theory, models, practice and algo-trading - page 3047

 
Aleksey Vyazmikin #:

It's just that the trouble is.

the trouble is something else entirely

 


 
Maxim Dmitrievsky #:


Victor Grigorievich, my respects xD

 
Aleksey Nikolayev #:

A purely theoretical question has arisen - can an ONNX model be used to derive another ONNX model. For example, the first model is used to periodically retrain on new data and update the working model. Meaning, without using python etc.

At first glance, it is unlikely to be possible, but in case someone has tried to do something like this.

I have not managed to get any meaningful answers from the AI - it writes that it can and cites references that have nothing to do with the question).

ONNX model is a graph of a trained model decomposed into elementary operations. It is impossible to train a model in ONNX format in Windows. They write about such a possibility in Linux.

It can be used only by getting a predicate (executes much faster than the model predicate) and without Python. Very interesting application of ONNX model in carefree-learn package. The picture below is taken from the package description.

carefree hiheline

 
mytarmailS #:

the problem is something else entirely.

Yes, found the cause.

In general, updated, even the error does not write, but the result is the same - all up almost.


And the previous code-script that was posted earlier stopped working - it used to work before the updates.

> p1 + p2 + plot_layout(nrow = 2)
Error in app$vspace(new_style$`margin-top` %||% 0) : 
  attempt to apply non-function
 
Aleksey Vyazmikin #:

Yeah, I found the cause.

In general updated, even the error does not write, but the result is the same - all up almost.


And the previous code-script that was posted earlier stopped working - it used to work before the updates.

Tomorrow I'll try to run it myself
 
mytarmailS #:

Anybody understandquanstrat package for creating, backtesting strategies of any level, optimising parameters, etc.

Created by real traders from the fund and is used every day for strategies with real money.

brief introduction

Some interesting thoughts from the same place

Rather than using backtests to validate good trading strategies, I think they are better suited to rejecting those strategies we definitely do NOT want to use.

====

how do i know my strategy is not overtrained or false in its returns?

  • For financial data, you can use upsampling or resampling (but you have to be very careful because of regime shifts).
  • Parameter sensitivity/stability testing. Basically, if I decide to use different parameters for my technical indicators (for example, SMA-30 and SMA-180) or start on different days of the week, will my results change a lot (how fragile are my indicators). model based on its initial conditions)?

well, if there are no other evaluation criteria, then through the stability of parameters

you can also imagine the output values of the TS as signals in time and measure their entropy and compare it with randomness. If the TC captures some patterns that repeat with some periodicity, this will be reflected.

For builders of custom FFs, it might be useful.

The best measure is time and tests in real life. Any TC will stop working.
 
Vladimir Perervenko #:

ONNX model is a graph of a trained model decomposed into elementary operations. It is impossible to train a model in ONNX format in Windows. They write about such a possibility in Linux.

It can be used only by getting a predicate (executes much faster than the model predicate) and without Python. Very interesting application of ONNX model in carefree-learn package. The picture below is taken from the package description.

The question about ONNX from ONNX arose simply from the juxtaposition of two statements I've encountered: 1) model acquisition can be represented as a pipeline, 2) the pipeline can be converted to ONNX format.

It is clear that this is hardly possible in practice. Actually, I would like to understand what exactly prevents the implementation of such a possibility to realise the fundamental limitations of this technology as a whole.

It's one thing if it's limitations like impossibility to write to a file and another if it's limitations like lack of support for data types (dataframes, for example).

 
Ivan Butko #:

Victor Grigorievich, my respects xD

Yeah, Especially today's video(((
 
Aleksey Vyazmikin #:

And the past code-script that was posted earlier a bit - stopped working - used to work - before the updates.

library(patchwork)

Is it installed?