Machine learning in trading: theory, models, practice and algo-trading - page 3426

New comment
 
mytarmailS #:

It's a long and complicated topic, in fact it copies the brain with its algorithms, and neural networks are not about the brain at all....


to summarise

Humans can transfer their experience to new knowledge that is different from the past, but neurons can't...

***

Well, it seems like multi-tasking and LLM are trying to solve these issues.

 
Maxim Dmitrievsky #:

Well, it seems like multi-tasking and LLM are trying to solve these issues.

No, it's not possible on current MO architectures and current hardware... but we're getting close...


quantum computers and processors with thousands of cores are appearing for a reason...

Theoretically, everything has been ready for a long time, only the hardware is lagging behind.

 
mytarmailS #:

No, it's not possible on current MoD architectures and current hardware... but we're getting close....


quantum computers and processors with thousands of cores... for a reason.

Theoretically, everything has been ready for a long time, only the hardware is lagging behind.

Here's a real graph and a generated one.

I'll put a bunch of such fractals into the model for processing, so to speak. Then I'll show you the results.


This is a double fractal. There can be up to three or four on the market, like this.

 
Maxim Dmitrievsky #:

Here's a look at the real graph and the generated one.

I'll put a bunch of such fractals into the model for processing, so to speak. I'll show you the results later.


What does this have to do with the Mandelbrot function?

I can do the same thing with randoms.

 
mytarmailS #:

and what does this have to do with the Mandelbrot function?

I can do the same thing with randoms.

It's not random, it's self-similar. I mean, it predicts well.

#27

that is, in fact, you're sticking a series that is more predictable, but that has similar patterns that are in forex. So you're probably improving your generalisation ability.

 
mytarmailS #:

and what does this have to do with the Mandelbrot function?

I can do the same thing with randoms.


library(quantmod)

n <- 50000

  rnorm(n, mean = 0.01, sd = 1) |> 
    cumsum() |> 
    xts(Sys.time()+1:n) |> 
    to.minutes5() |> 
    chart_Series()


 
mytarmailS #:

Yours is not predictable.

 
Maxim Dmitrievsky #:

it's not random, it's self-similar. So it's a good predictor.

#27

Is there any confirmation that the market corresponds to this function?

 
mytarmailS #:

Is there any confirmation that the market matches this function?

I've been into this for a long time. There are books:

Disobedient Markets. Mandelbrot.

Fractal theory of the forex market. Almazov.

In general, the analogy can be traced.

 
Maxim Dmitrievsky #:

Python has a tree in lib sklearn, which is almost all fully convertible to onnx

I don't need just a tree, I need consistent clustering. Like in the diagram below. The number of clusters at each level should be set arbitrarily, as well as the depth of the tree.

Maxim Dmitrievsky #:

and chatgpt knows python well

I still don't understand what is keeping you.

I use it, and it is sometimes very dumb.... I.e. relies on old syntax in packages, even though they have changed before learning. Apparently it's because there is less correct data than incorrect data.

Otherwise, it's a good tool!

1...341934203421342234233424342534263427342834293430343134323433...3548
New comment
Reason: