Machine learning in trading: theory, models, practice and algo-trading - page 2038
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
there is nothing to argue about, because in any normal framework did and showed, with a minimum of code
i don't particularly discuss self-made models here, only mature models like catbust or modern neural networks
It's not even interesting to discuss this mouse fuss with mql neural networks, because the world is far ahead, and every year it doubles the gap.
Suppose you tell me: "I have such-and-such a model on tensorflow"... I say "good, I can make the same model on a torch for 5 minutes and check it. And you tell me that you dug something in mql. What do I need this information for? How can I recreate it?
Why do not you admit that I have a model thattensorfloat / torch a couple of years, and at the same time I triple the gap, why do not you assume?)
from some book by Edgar Peters, not verbatim of course - "The clerk/broker will pass a five pound bill lying on some street because in his probability theory it shouldn't be there...")
Good luck.
The more lines, the more depth is needed.
If there are gigabytes, then millions of lines. At depth 6, the final sheet will be 1/64th of the full number of examples/strings, i.e. tens of thousands if the input is millions.
So CB builds ensembles, i.e. cutting happens both ways, and the minutiae depends on the number of trees in general they recommend a maximum depth of 10.
Why do not you admit that I have such a model to whichthe tensorfloo/torchu another couple of years, and at the same time I triple the gap, why do not you allow?)
it's already downhill talk, unfortunately
So CB builds ensembles, i.e. cutting happens and so, and mincing depends on the number of trees in general they recommend a maximum depth of 10.
Apparently this is not a recommendation for millions of lines. Try 15.
Apparently this is not a recommendation for millions of lines. Try 15.
This is just for very large samples in width and depth.
I'm saying that the training is easy, but the patterns detected by the training are not confirmed on the control sample.
This is just for very large samples in width and depth.
I'm saying that the training is easy, but the identified patterns in the training are not confirmed in the control sample.
There is no problem with the analysis of market information... except for greed of a researcher who believes that the market gives information only to him and he needs to process all data, i.e. here the problem is formalized as we are looking for a repeating pattern, other data should be discarded (not used)
With the decision is sad - to generate TS that will pass the test and forward is possible, but to find links between the strategy tester statistics and the lifetime of the TS or the possibility of determining the relevance of the TS to the market context - that's the problem
i.e., as you write the problem is in the future
I think that in general we have made a little progress in the formalization of the problem,
In principle it is not difficult to make an unloading of testing statistics and try to teach TC in Python,
Determining market context, imho it's like you wrote - only a trader's decision, i.e. i doubt it can be formalized or algorithmized or investigated
With atoms it is simpler 2 levels matter molecules atoms in inorganics and matter large molecules and lots of atoms in organics. In state market analysis, business entities are people. and people are more complicated than atoms and their constituents. definitely more complicated than organic chemistry.
In my opinion, the number of columns is not so important for depth. Only a large number of rows should increase the depth. As for the columns, you can have as many as a million. They will all be checked and only some of the best ones will be chosen.
In general, this is the result - 2 trees in the end
The metric
Evaluation of prefactors importance according to the selected method
I have attached C++ and bin model and quantization table.
Seems to me that either the sample should be mixed, or this is the maximum - confused by the difference in metrics between the test and exam samples.
With atoms it is easier to have two levels of matter molecules atoms in inorganics and matter large molecules and lots of atoms in organics.
This is exactly the case, but as long as you consider atoms, and as soon as you decide to consider the crystal lattice, "it's all over" - in most cases a very difficult problem to solve, I read once, by the way they use neural networks - faster results of the modeling process are obtained
It is the same with the markets, while we look at one or two ticks, there are not many variants - either up or down, everything becomes complicated if we go to the study of "here they opened an order, and here they closed it" )))
In market analysis of states, business entities are people. and people are more complicated than atoms and their constituents. definitely more complicated than organic chemistry.
Yes, this is a living structure, but it has a behavioral model - let it be called patterns or the market context, not the point.
It is possible to study the behavior on the history, and now it's up to you to decide what to do in the future.
In general, this is the result - 2 trees in the end
Metrics
Estimation of importance of predictors according to selected method
I have attached C++ and bin model and quantization table.
Seems to me that the sample either needs to be shuffled, or this is the maximum - confused by the difference in metrics between the test and exam samples.
You can mix inside the train or test, but it makes no sense, and you can't mix between train and test. Did you mix them there by any chance, something very good result for a test with an exam.
The difference is that the market is not always the same; it changes. The main thing is the average profit).