Principles of working with an optimiser and basic ways of avoiding fitting in. - page 9
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
No need to despair as the ARTICLE has already been written.
Thanks for the interesting material. But still, it's more of a technical article. I'm more interested in conceptual questions, not related to any particular trading discipline. I'll continue the discussion in the stated vein a bit later.
Read the article. I can't agree with everything, but a lot of sensible and most importantly correct thoughts.
No common sense, no right thinking. The usual nerdish game of numbers. Here is a quote from the beginning of the article:
The input parameters are unrelated to the output parameters, i.e. there is no causal relationship between them - they are unrepresentative, and therefore there is nothing to train the neural network. And checking for correlations between inputs and outputs will show a correlation close to zero. In this case you need to look for other input data to train the neuronet on
A person doesn't deal with structuring of input data and has to write "we must search". Look for what? Why aren't they correlated? Because you took an idiotic indicator? Do we need to find a less retarded indicator? There is no answer to the original questions, so Tarsky and Gedel and arithmetic are just nerdy nerdy.
The article is harmful because there is a simple truth behind the science: all that is written is true only for the kotir for which the experiment was performed. There is no reason to trust any extrapolation of these results. All that is written is a simple breakpoint test. But apart from this test people know that a bunch of additional tests need to be done, which will increase confidence in the future extrapolation, but still won't solve the problem.
Reshetov! You need to be more modest. Nothing new in the problem of ensuring robustness. Just another invention of the bicycle. But what a mind game!
faa1947:
No common sense, no correct thoughts. The usual nerdish game of numbers.
...
The article is damaging because the simple truth is hiding behind the science
...
For the sake of objectivity, your review has been posted in the comments of the article with a link to the source.
It is not a review, but an attempt to move you to read specialized literature.
Previously, Gödel's theorem was read as part of a philosophy course called "Theory of Knowledge" - a certain kind of stovepiped, gives some framework in reasoning and nothing more. We have concrete economics and there are some, rather modest, advances in it. The breakpoint test is the first test in a series of stability tests. And if we dance, we should dance with the solutions that this kind of test gives, which compares the results for two quotient areas.
But besides this kind of tests, there is another set of tests that checks the model for other reasons of instability. What is the place of what you have written among this set of solved problems? What solves your proposals among what is currently unsolved?
No common sense, no right thinking. The usual nerdy game of numbers.
It's not a review, it's an attempt to move you to read specialist literature.
It's no use, I'm a Chukcha the writer. I never learned how to read, and immediately started writing harmful articles.
So, all hope is on you as a writer of useful articles.
The input parameters are unrelated to the output parameters, i.e. there is no causal relationship between them - they are unrepresentative and therefore there is nothing to train the neural network. And checking for correlations between inputs and outputs will show a correlation close to zero. In this case you should look for other input data to train the neuronet on
A person doesn't deal with structuring of input data and has to write "we must search". Look for what? Why aren't they correlated? Because you took an idiotic indicator? Do we need to find a less retarded indicator? There is no answer to the original questions, so Tarsky and Gedel and arithmetic are just nerdy blather.
In fact - absolutely correct and valid point. If there is no relationship of any kind between the input data and the data on which the network is trained, then trying to get a profit in such circumstances is absolutely useless.
faa1947 : This is not a review, but an attempt to move you to read specialist literature.
In fact - absolutely correct and valid point. If there is no correlation between the input data and the data on which the network is trained, then trying to get a profit in such conditions is absolutely useless.
It's a trivial thought like: we're wet, so the boots are flowing. And where do we get the boots? We don't know. It has to be in the kodobase. That's the problem.
And where have you seen specialized literature on "how to avoid fitting or retraining a neural network" or "how to make a TS with a neural network and trade in financial markets"?
This is a common problem and they are trying to solve it, and the use of NS is not a reason to ignore existing developments.
faa1947: Это банальная мысль типа: мы промокли, значит сапоги текут. А где взять сапоги? Не известно. Должно быть в кодобазе. Проблема в этом.
I don't see the correlation between the input data and the data on which the network learns and the boots...)))), which also flows for unknown reasons....))))
faa1947 : This is a common problem and they are trying to solve it, and the use of NS is not a reason to ignore existing developments.
These are general words, and using them is no reason to applaud someone else's article, on which a person has worked and tried to convey something useful to people.IMHO it is pointless to exhort. Mr Economist was not satisfied with the article. So what the hell with them. There's no arguing about tastes.
Whoever has found useful information is already using it. Whoever hasn't found anything for themselves and is dissatisfied - that's not my problem.