From theory to practice - page 276
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
That's a bit closer to the subject.)) However, count it as you like.)
I'm not even going to get into these distributions. The distribution is what it is in reality, and attempts to fit it to something with a name are, imho, unfounded. Why should it correspond to something specific that is already known?
Say, nobody even tried to describe the distribution of black body radiation by already known distributions. Why the hell are we trying to match something already known here?
Yuri, stop dispelling the grail! This is an outrage! We almost believe it!)
I'm not dispelling it at all.
Once, a long time ago, I attended a seminar at the Keldysh Institute. I no longer remember anything, not even the topics of the lectures. However, there was one very interesting idea - the more complex the system, the simpler the model should be. That is, simpler models give the most accurate descriptions, within reason, of course.
I'm not dispelling it at all.
Once, a long time ago, I attended a seminar at the Keldysh Institute. I no longer remember anything, not even the topics of the lectures. However, there was one very interesting idea - the more complex the system, the simpler the model should be. That is, simpler models give the most accurate descriptions, within reason, of course.
there you go!
exactly
I'm not dispelling it at all.
Once, a long time ago, I attended a seminar at the Keldysh Institute. I no longer remember anything, not even the topics of the lectures. However, there was one very interesting idea - the more complex the system, the simpler the model should be. That is, simpler models give the most accurate descriptions, within reason, of course.
The thought is certainly an interesting one. That speaker was probably from a group studying the behaviour of a spherical horse in a vacuum))
IMHO, the model, first of all, should be adequate. That is, it should reflect the processes taking place inside the object. And the more accurate, the better (this is in accordance with capability and ability).
Then it will be viable and practically useful.
As always, I may be mistaken.
What you're driving at))). What, well what-what. You don't have a working one, either simple or elementary. Anybody scratching their head, it's like, well-I knew it, I agree. Sectarians.
apchi
I was going to add there, not even like that, but like this:
a simple system, a simple model.
simple system, elementary model.
forex is a simple system, 100%
I practically proved for myself that a model with less inputs and polynomial length is smaller than a model with more inputs and polynomial length is usually more adequate for the market, which does not fit the logic according to which the more complex the model is, the smarter it is. Of course, this effect is not always true and sometimes a very small model is inadequate as well. But... I have found a way to choose a model that is the most adequate of all presented models, at that on the training area without wasting a precious plot of OOS.
Imagine you received several models and after evaluation with a fairly high accuracy chose the one that will score in the future. And it dials in...... That was my breakthrough just under a month ago...
Prove it to your deposit.
The thought, of course, is interesting. That speaker must have been from a group studying the behaviour of the spherical horse in a vacuum.)
IMHO, a model, first of all, should be adequate. That is, it should reflect the processes taking place inside the object. And the more accurate, the better (this is in accordance with capability and ability).
Then it will be viable and practically useful.
As always, I may be mistaken.
Let's start with the black box - we don't know what's going on inside it. What - "reflect the processes going on inside the object" can we be talking about? And the question about the accuracy of the description of the processes in the object is wrong. A model of the BS is not meant to describe processes inside the BS at all. The model must describe the behaviour of the system as a whole.
The requirement of simplicity just gives viability, and the complication gives excellent convergence only in the model development section. You can show this on simple regression models, where simple describes the process more adequately.
Yes, and simplicity should not be confused with primitiveness.
It's just more logical to talk about quantitative estimates. And the benefits of complex systems are so insignificant that when it comes to quality/results, they lose out to simple ones. It's like if you take the system that guesses the target closest, and a simple one that guesses the target less accurately. But there are both positive and negative trades. More accurate (fat) plus will give more accurate (fat) minus. So there is no sense in complicating it. Besides, if a complex model is used in iterative methods, then at any complication resources will be spent in seven-mile steps. And time too. So the complication also depends on specific conditions. If output is 3 kopecks, but resources eat up to a lakh, is it worth it? This is not because simple is more accurate. It is because the accuracy of a complex one is less significant in total.
You are saying all the right things. And that's all important too. But I was writing about a slightly different thing. About the increase in model error when the model becomes more complex, say over some threshold. For example, for a Wiener process, the best predictor is the current value. Trying to make the model more complex leads to a decrease in prediction accuracy, and a simpler model is preferable. When modelling other systems, it's generally the same.
It is simply more logical to talk about quantitative assessments. And the benefits of complex systems are so insignificant that when choosing quality/outcome-lose out to simple ones. It's like if you take the system that guesses the target closest, and a simple one that guesses the target less accurately. But there are positive and negative deals. More accurate (fat) plus will give more accurate (fat) minus. So there is no sense in complicating it. Besides, if a complex model is used in iterative methods, resources will be spent in seven-mile steps at any slightest complication. And time too. So the complication also depends on specific conditions. If output is 3 kopecks, but resources eat up to a lakh, is it worth it? This is not because simple is more accurate. It's because the accuracy of a complex one is less significant in total.