Discussing the article: "Biological neuron for forecasting financial time series"

New comment
 

Check out the new article: Biological neuron for forecasting financial time series.

We will build a biologically correct system of neurons for time series forecasting. The introduction of a plasma-like environment into the neural network architecture creates a kind of "collective intelligence," where each neuron influences the system's operation not only through direct connections, but also through long-range electromagnetic interactions. Let's see how the neural brain modeling system will perform in the market.

The Nobel Prize-winning Hodgkin-Huxley model describes the mechanism of generation and propagation of nerve impulses at the cellular level. But why might this particular model be the key to understanding financial markets? The answer lies in the striking analogy between the spread of nerve impulses in the brain and the spread of information in markets. Just as neurons exchange electrical signals through synaptic connections, market participants exchange information through trading transactions.

The innovation of our approach lies in adding a plasma-like component to the classical model. We view a neural network as a dynamic system immersed in the "plasma" of market information, where each neuron can influence the behavior of other neurons not only through direct connections, but also through the electromagnetic fields it creates. This allows the system to capture subtle correlations and relationships that go unnoticed by traditional algorithms.

In this article, we will take a detailed look at the system architecture, its operating principles, and the results of practical application on various financial instruments. We will show how a biologically inspired approach can offer a new perspective on the problem of financial time series forecasting and open new horizons in the field of algorithmic trading.


Author: Yevgeniy Koshtenko

 
Neural networks are still a subject of study for me, I plan to use them in my scalper. I do not see any mysteries in them. For me, the mystery is in another - why do authors of such articles try to feed raw bars to NS with the persistence of a maniac? I think that if a person has mastered the work with NS, then the basics of DSP (digital signal processing) will not be difficult to learn. In this article, the author has surpassed all the so-called "analysts" I know - he feeds D1 bars to the input and tries to guess the price 15 days ahead. Is it so hard to extract tick data from MT5 and try it on scalping with preliminary processing????
 
Alexey Volchanskiy #:
Neural networks are still a subject of study for me, I plan to use them in my scalper. I do not see any mysteries in them. For me, the mystery is in another - why do authors of such articles try to feed raw bars to NS with the persistence of a maniac? I think that if a person has mastered the work with NS, then the basics of DSP (digital signal processing) will not be difficult to learn. In this article, the author has surpassed all the so-called "analysts" I know - he feeds D1 bars to the input and tries to guess the price 15 days ahead. Is it so hard to extract tick data from MT5 and try it on scalping with pre-processing????
Is pre-processing like going to a vanga?

How do you know that prices need to be processed, where is the justification for it, except for the slogans of MO-schniks about the presence of phantom noise, the definition of which in the context of forex is not even known.

Price is not a physical signal, there is no noise.

Each price pattern describes its own trend and flatness, whether on D1 or M1.
Impulse-correction.

All TFs have their own patterns.

Chart preprocessing is the search for patterns. Searching for patterns through all kinds of filters.

That's what any NS does.

You'll just be processing the chart twice.

And if you input indicators - 3 times.
 

To Ivan Butko

Preprocessing - preprocessing of predictors - is the first and most important of the three stages of any machine learning project. You need to sit down and learn the basics. Then you wouldn't be talking rubbish.

"Garbage in - rubbish out" - and you don't need to go to a fortune-teller for that.

 

From the article;

Exotics offer no advantage even against simple statistical models. And for what?

By code:

Adaptive normalisation - I didn't see what is adaptive there?

All indicators are in the library of technical analysis ta. Why rewrite everything in Python?

There is no sense in practical application, IMHO

 
Vladimir Perervenko project. You need to sit down and learn the basics. Then you would not be talking rubbish.

"Garbage in - rubbish out" - and you don't need to go to a fortune-teller for that.

You translate textbooks

You haven't dealt with the definition of rubbish in prices

You don't know what is rubbish and what is not. And whether it exists in principle. Since at Forex people earn on M1, and on M5, and on M15 and so on, up to D1

You do not understand and do not know how to trade with hands.

Hence - you do not understand what you yourself are saying.


But if you have a confirmation of workability and stability of your NS models solely because of the presence of preprocessing (without it - rubbish) - you will be right.

Are there such?
 
Easy to say: learn the basics, it's at least 1 book on the basics you need to read :) and not just read, but memorise.
 
Noise is model error. That is, in reality there is no abstract "price noise", there is only a number of errors of a particular model. A model is considered more or less working if a number of these errors behaves as white noise (stationary process without correlation).
New comment