Machine learning in trading: theory, models, practice and algo-trading - page 2196

 
Maxim Dmitrievsky:

Finished a course on rollups, although I gave up on it, because it is not interesting just to stick in the videos on YouTube

But, now there are a lot of free paid courses, such as from coursera. I'll take another one )). For some reason they recommend English. But it makes no difference. The more courses you take, the higher your level. Then they give you a digital certificate.


Where do you pass?
 
Alexander Alekseevich:
Where do you pass?

https://2035.university/

Университет 20.35
Университет 20.35
  • 2035.university
Первый в России университет, обеспечивающий профессиональное развитие человека в цифровой экономике. Готовит лидеров компаний, участников Национальной технологической инициативы.
 
Maxim Dmitrievsky:

Yeah, cool. My tensor for some reason does not want to stand up, so I will look at the torch

I'm not even so much interested in the implementation as in the approach itself. How to set distributions, interpret, etc. Because it is possible to do in different ways

here are interesting articles 1, 2, give an intuition of what's going on inside

Cool!

Here are examples of various generative models including VAE with examples for pytorch, might be useful

https://github.com/wiseodd/generative-models

wiseodd/generative-models
wiseodd/generative-models
  • wiseodd
  • github.com
Collection of generative models, e.g. GAN, VAE in Pytorch and Tensorflow. Also present here are RBM and Helmholtz Machine. Note: Generated samples will be stored in (or , etc) directory during training. What's in it? Generative Adversarial Nets (GAN) Vanilla GAN Conditional GAN InfoGAN Wasserstein GAN Mode Regularized GAN Coupled GAN Auxiliary...
 

these levels, it's so complicated...

Tell me, who would have thought that this section on the left could be the reason for this bounce on the right?

And how much extra information in between?

I'll brag a little and say that I suspected or was almost sure that it was so....

But this is not the main thing, the main thing is that there is no "out of the box" algorithm that could find such relationships .....

I searched a lot of scientific literature in search of solutions for similar problem with the market, but apparently there is no such a similar problem...

That's why there are no algorithms able to solve this problem...

And by the way, looking at the picture, you can understand that the market is not a time series, it's like written in a time series format, but you have to work with it in a different way, with understanding of its structure ...

That means that all known approaches don't work, which in principle is confirmed by practice...


I'm not advocating anything and I'm not trying to change anyone's mind ... I just wanted to write and wrote...

 
mytarmailS:

these levels, it's so complicated...

Tell me, who would have thought that this section on the left could be the reason for this bounce on the right?

And how much extra information in between?

I'll brag a little and say that I suspected or was almost sure that it was so....

But this is not the main thing, the main thing is that there is no "out of the box" algorithm, which would be able to find such relationships .....

I searched a lot of scientific literature in search of solutions for similar problem with the market, but apparently there is no such a similar problem...

That's why there are no algorithms able to solve this problem...

And by the way, looking at the picture, you can understand that the market is not a time series, it's like written in a time series format, but you have to work with it in a different way, with understanding of its structure ...

That means that all known approaches don't work, which in principle is confirmed by practice...


I'm not advocating anything and I'm not trying to change anyone's mind ... I just wanted to write and wrote...

It is necessary to confirm the hypothesis. It may well be. But it doesn't make sense. We have to look at different sites.

 
Valeriy Yastremskiy:

We have to confirm the hypothesis. It may well be. But it doesn't seem logical. We have to look at different sections.

What is not logical?

 
mytarmailS:

these levels, it's so complicated...

Tell me, who would have thought that this section on the left could be the reason for this bounce on the right?

And how much extra information in between?

I'll boast a little and say that I suspected or was almost sure that it was so....

But this is not the main thing, the main thing is that there is no "out of the box" algorithm that could find such relationships .....

I searched a lot of scientific literature in search of solutions for a similar problem with the market, but apparently there is no such a similar problem...

That's why there are no algorithms able to solve this problem...

And by the way, looking at the picture, you can understand that the market is not a time series, it's like written in a time series format, but you have to work with it in a different way, with understanding of its structure ...

That means that all known approaches don't work, which in principle is confirmed by practice...


I'm not advocating anything and I'm not trying to change anyone's mind ... I just wanted to write and wrote...

You are trying to predict too far ahead. 5 to 10 bars ahead is still possible. But 100 is unrealistic.
 
mytarmailS:

What's not logical?

That the pattern is repetitive and not random. Although not randomness is secondary.

 
elibrarius:
You are trying to predict too far ahead. 5 to 10 bars ahead is still possible. But for 100 it's unreal.

The matter is that I'm not forecasting the pattern, but the rebounce price, and the necessary rebounce may happen in 5 candles or in 55, it's the same pattern.

This is what I am saying, the approaches to the market as to BP do not work, because it is the prices (levels) that are more important on the market than the waves of price movements (patterns)

 
mytarmailS:

The thing is that I'm not forecasting a pattern but a rebound price, and the rebound at the right price may happen after 5 candles or after 55 candles, it doesn't matter, and it will be the same pattern...

That's what I'm saying, the approach to the market as a BP doesn't work, because it's the prices (levels) that are important in the market, and not the waves of price movements (patterns).

The farther the forecast, the less accurate it is. A lot can happen in 100 bars, an order of magnitude more than in 10.