Machine learning in trading: theory, models, practice and algo-trading - page 1797
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
Thought about it (we're talking about some place to discuss the application of theorist and matstat). I believe that nowadays it is better to do something like this in the form of articles and their subsequent discussion.
Shared this idea with the administration and got preliminary approval. If you have any ideas and suggestions for article topics, then share them here.
Science is good if it saves you time and effort. Here I have another crazy idea - to collect new leaves based on splits of other leaves, the question is, what is the probability that this will have a positive effect and there will be new leaves that have little correlation with the former ones?
As I think, this approach, unlike the greedy principle, will allow to reduce the completeness of the sample more smoothly, and therefore should make the signal from the leaf more stable.
Toddler posted a fresh piece. Haven't read it yet.
Apparently, another modernization of the perpetual motion machine, increased cross-country capability.
Thought about it (we're talking about some place to discuss the application of theorist and matstat). I believe that nowadays it is better to do something like this in the form of articles and their subsequent discussion.
Shared this idea with the administration and got preliminary approval. If there are any ideas and suggestions for article topics, please share them here.
Really difficult to choose topics from complex science and what would be useful for people)))
If on directions, apparently practical application of various models in treatment of BP. Parsing of various NS algorithms for BP with linkage to mathematical models. Probabilistic expectation of NS results. Theorver and MO algorithms, which are better, which are worse).
I like the latter.
Thanks for the thought to read.)) There's some lucid material on the matstat, too.)) The conclusion that the series can be estimated / defined / characterized by the model is there and as a topic definition of the minimum area to determine a reliable model.
There are closed invitation groups here on the site
But better articles
Will the NS read minds? Link.
I'm dealing with arma. While getting to the bottom of it, I learned a lot of interesting things. I always thought that unit root test is somehow related to imaginary unit. It turned out to be a test for filter stability. arma is a filter consisting of bi and k filters. The filter can be represented as a ratio of B/A polynomials. Their roots are called zeros and poles. Their position on the unit circle is used to judge the stability of the filter. The unit root test is just about this.
I can not make a stable arma filter, with ar there is no such a problem.
I'm dealing with arma. While getting to the bottom of it, I learned a lot of interesting things. I always thought that unit root test is somehow related to imaginary unit. It turned out to be a test for filter stability. arma is a filter consisting of bich and ki filters. The filter can be represented as a ratio of B/A polynomials. Their roots are called zeros and poles. Their position on the unit circle is used to judge the stability of the filter. The unit root test is just about this.
No way to make a stable arma filter, with ar there is no such a problem.
You can say that the filter, you can say that it is a method describing BP. If the model unstably describes BP, then we either have to increase the number of polynomials or move to a more complex or different model.
Will the NS read minds? Link.
The work was even before NS and MO. The results in pattern recognition were there long ago, but they were not written about in this way. AI, NS, MO are just drivers of advertising and promotion)))) An AI mop is always more expensive than a simple mop.))
As I think, this approach, unlike the principle of greed, will allow to reduce the completeness of sampling more smoothly, and therefore should make the signal from the leaf more stable.
it is better to direct the energy to the generation of new features... IMHO...
If the features are good, then we can even try to predict extrema, I have about 400 and 10 models, but I see that I need tens of times more, training samples can reach > 100 GB , no way )))) I don't have such power ...
From the pictures you can see that the algorithm is able to recognize something, but we need to build up the features...
The work was done even before the NS and the MO. The results in pattern recognition were long ago, but they were not written about them in this way. AI, NS, MO are just advertising and promotion drivers)))) An AI mop is always more expensive than a simple mop)))
Yeah, sure )))) I would like to see how they recognize images or generate speech with "if, then, else" that would be hardcore)) But seriously, you're talking nonsense!