Machine learning in trading: theory, models, practice and algo-trading - page 3737
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, the founder was Lamarck.
Lamarck's theory of evolution was indeed the first, but it was crude and had hypotheses rejected later (inheritance of traits). Nowadays, Darwin's theory of evolution with natural selection is still considered dominant (although it is also not proved by anything, and, on the contrary, there is a lot of evidence against it).
Informativeness and relevance (highlighting the latter areas of structures of different scales) is a subject of study both with one's own brain and with the help of ME.
I write: to go deep into the study of the structures of the SOURCE data.
At the same time, I suggest to engage in self-education - on the topics: space, geometry, forces, symmetry.
Judging by the reaction, all these words are like Chinese for the participants.
I write: to go deep into the study of the structures of the SOURCE data.
At the same time, I suggest to engage in self-education - on the topics: space, geometry, forces, symmetry.
Judging by the reaction, all these words are like Chinese for the participants.
I write: delve into the study of the structures of the source data.
Ivan Butko #:
And models (most of them probably) have adders, so they will react to the value of a number.
Imagine that there is a way to make the models not care about the magnitude of the number, and only the fact that the numbers are different will be important.
It's called categorical features. They don't care: red, warm, 0.9 or 0.1. You just have to specify that the feature is not numerical, but categorical. I'm not sure about neural networks (probably not), but in wooden models it is easily realised. Among the well-known ones - Catboost has it.
Apparently, the textbooks and instructions for MO models were read through the line....
I am not a master of MO, but from my small experience I see it this way: it is possible to study price movements to find patterns with MO, but it is not enough for trading. The system tries to describe the market, but in isolation from the real deal. A deal goes through several stages in its "life". Where to open a deal, what are the grounds for it? Then, a deal, for example, is in a plus, where is the profit limit, what are the reasons to close it or continue saving profit? Or a trade in a loss and we have a favourable pullback, to exit with a small minus or wait for a plus, what are the odds? We may be right in our entry, or we may be wrong, and we should look for solutions based on our specific circumstances, not just general market patterns, until we close this trade. That is, we should teach the neural network not regularities, but behavioural strategy. Without pretensions to super knowledge, it seems to make sense
Mode detection - not sure what that is, sounds interesting. Perhaps it works. Sometimes, I encounter that in the current losing trade, I ignore the counter signal, for other indirect reasons. As a result, it turns out to close the trade with a profit, instead of fussing with two losing trades. This is to the fact that at the moment of opening - one information, in the course of events development is already different. And to open a deal, put a take, stop and wait for the result - a bit flat strategy training for real trading.