Machine learning in trading: theory, models, practice and algo-trading - page 3086
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
Friends, hello!
There is a battle, welcome, make some noise!!!
here is a losing option from the beginning, because you have a lot of them up your sleeve, and we do binary classification the old-fashioned way :)
and not every function can be painlessly fed to a neural network.
it's a lose-lose here, because you have a lot of them up your sleeve, and we're doing binary classification the old-fashioned way :)
and not every function can be painlessly fed to a neural network.
Actually, a million parameters is the "great equaliser", the search space is so large that I don't know which algorithm will be the winner. And what will be in the black box is unknown (or rather, it is known, but we need to find the "key").
It's fun, it's like cracking a safe!
In fact, a million parameters is the "great equaliser", the search space is so large that I don't know which algorithm will be the winner. And what will be in the black box is unknown (or rather, it is known, but it is necessary to find the "key").
It's fun, it's like cracking a safe!
and not every function can be painlessly fed to a neural net.
a couple of three years ago, a friend from faraway Australia approached me and said, let's mine bitcoin with an algorithm! make me a thing that can find the next block. people were generating the 7th digit in the hash, our algo was able to find up to the 5th digit in less than an hour.... we were too late.
And so it is here. someone will try to participate in the championship for the sake of interest, he will come up with new bright ideas, and it will be useful for him.
In fact, a million parameters is the "great equaliser", the search space is so large that I don't know which algorithm will be the winner. And what will be in the black box is unknown (or rather, it is known, but it is necessary to find the "key").
It's fun, it's like cracking a safe!
It's not a matter of algorithm, it's a matter of genetics, or swarming, or whatever.
1) It's a question of time and power of iron !!!! who has more time and stronger iron will win.
2) the obtained results will in no way guarantee that this particular AO is the best, as the best AO is likely to become the best by chance (it just happened to find the best maximum).
3) more than 20-30 measurements in a function is already a guessing game and in real problems nobody works with AO on such huge measurements of a million parameters (measurements reduce).
4) the problem itself is constructed incorrectly, it does not reveal the peculiarities of AO in any way, everything is built on - whoever is lucky enough to find the best maximum wins.
To find the maximum in the function for 10 iterations is a normal problem, which will reveal the efficiency of AO, and this is how problems are set in normal circles ...
But what good is it when you talk to a profane who thinks he is an expert, and his friend and adviser is gpt chat )))
It's not a matter of algorithm, it's a matter of genetics, swarming, whatever.
1) It's a question of time and hardware power !!!! who has more time and stronger hardware will win.
2) the obtained results will not guarantee that this particular AO is the best, as the best AO will become the best most likely by chance (it just happened that he found the best maximum).
3) more than 20-30 measurements in a function is already a guessing game and in real tasks nobody works with AO on such huge measurements of a million parameters (measurements reduce).
4) the problem itself is constructed incorrectly, it does not reveal the peculiarities of AO in any way, everything is built on - whoever is lucky enough to find the best maximum wins.
5) Finding the maximum in the function for 10 iterations is a normal problem, which will reveal the efficiency of AO, and this is how problems are set in normal circles ...
6) But what good is it when you talk to a profane person who thinks he is an expert, and his friend and adviser is gpt chat )))
1. Black box is impossible to run more than 10000, this has been voiced. No matter how powerful hardware is used - it won't help.
2. You can't get a non-random result by chance on a million parameters, in a random search the results are averaged. The chance to find a result better than others only if the algorithm is better. To understand this you need to know a little probability theory or at least have some analytical skills.
3. In real problems there are billions of variables - modern generative networks. There are several billion neurons in the human brain and you have to learn every day to understand what we are talking about here.
4. You will not be lucky, I give you 100%.
5. Stochastic algorithms start with random numbers within an acceptable range, the fewer iterations, the more random the result. Additionally see point 2.
6. not in vain you are glued - pateushnik..... militant ignorance.
New features in MQL5:
Friends, hello!
There is a battle, welcome, make some noise!!!
Long live the collective farm!
Let's shame the professionals!
https://cran.r-project.org/web/views/Optimization.html
Long live the collective farm!
Let's shame the professionals!
https://cran.r-project.org/web/views/Optimization.html
Come by all means, since there is integration of R with MT5, it will be interesting how good the tools are.
No one needs to shame anyone, the one who solves the problem better is the best.