Discussing the article: "Role of random number generator quality in the efficiency of optimization algorithms" - page 14

 
Andrey Dik #:

I am in no mood to engage in dialogue in a disrespectful manner.

There was a championship of optimisation algorithms, where there was an opportunity to show your knowledge and skills, but somehow I didn't see people drawing poop there now.

Really?

Really?

And then Dick's pointless and merciless rubbish about completely ignoring my posts started. You just wandered into the thread and started talking rubbish.

I didn't ban you, I didn't complain about you, I let you freely carry your verbal diarrhoea. What's the problem now that I've pointed out your blunders?

Машинное обучение в трейдинге: теория, модели, практика и алготорговля - Находитесь в обществе таких, про которых вы практически ничего не знаете.
Машинное обучение в трейдинге: теория, модели, практика и алготорговля - Находитесь в обществе таких, про которых вы практически ничего не знаете.
  • 2023.10.26
  • Renat Akhtyamov
  • www.mql5.com
Тебе собственное абстрактное мышление не подсказывает что означает слово. Здесь нужно уметь уступать и признавать свои ошибки. про которых ты практически ничего не знаешь, которые могут знать в разы больше тебя
 
Yuriy Bykov #:

Probably just a very similar experiment just conducted in the last article. Indeed, the best results on the forward period were those sets that performed well but not the best on the main period.

Another example. Let's consider a monosymbol TS - it trades exactly on one symbol (in this case trading Signals can be formed on the basis of analysing prices/events of other symbols).

One of the input parameters of such a TS - on which symbol we trade. Let the range of this parameter be 10 symbols. Somehow we know that only one of them (X) is working.


We include this parameter in the parameter to be optimised and the output is one symbol Y != X. For some reason it's almost obvious, that's why you don't do this - optimise each symbol separately.


By "symbol" you can consider any input parameter. Local vertices - useful information for TC analysis.

 
fxsaber #:

It would probably be convenient to have a browser extension for each MQL5 branch, where you could ask not to show certain posts. I would like to see only constructive posts.

And where did you see nonconstructive behaviour on my part?
Proving that a problem has no solution is also a solution.
But it seems that you do not believe in the lack of a solution, so you see the lack of constructive behaviour
Then I'm out of here. I've done all I can.
 
Nikolai Semko #:
Where did you see anything unconstructive on my part?

Speculation.

 
fxsaber #:

Another example. Let's consider a monosymbol TS - trades exactly on one symbol (in this case trading Signals can be formed based on the analysis of prices/events of other symbols).

One of the input parameters of such a TS - on which symbol we trade. Let the range of this parameter be 10 symbols. Somehow we know that only one of them (X) is working.

We include this parameter in the parameter to be optimised and the output is one symbol Y != X. For some reason it is almost obvious, so we don't do it this way - optimise each symbol separately.

By "symbol" you can consider any input parameter. Local vertices - useful information for TC analysis.

There seems to be a contradiction here: if we know in advance that only symbol X is good, then no optimisation should find another symbol Y. And if it does, it means that our initial information that only X is good is incorrect. From a false premise we can get both true and false conclusions.

It seems to me that each symbol is optimised separately not because it is impossible to find optimal combinations or to find some "wrong" ones. From the point of view of the optimisation algorithm, it doesn't care what meaning we put into the parameter values being searched. Probably, by optimising by individual symbols we just help the algorithm to find optimal values faster. Or even very much faster.

 
Yuriy Bykov #:

There seems to be a contradiction here: if we know in advance that only symbol X is good, then no optimisation should find another symbol Y. And if it does, then our initial information that only X is good is wrong. From a false premise we can get both true and false conclusions.

We "know" in order to show that Y != X. If we restate that we assume that there is one working symbol. But the optimisation does not point to it. Thus if we take all local vertices, they will contain all brute-force symbols.

It seems to me that each symbol is optimised separately not because it is impossible to find optimal combinations or to find some "wrong" ones. From the point of view of the optimisation algorithm, it doesn't care what meaning we put into the parameter values being searched. Probably, by optimising by individual symbols we just help the algorithm to find optimal values faster. Or even very much faster.

If the optimisation would calculate instantly, I wouldn't put a symbol in the parameter being optimised.

 
Unfortunately, it became even less clear what it was about.
 
Yuriy Bykov #:
Unfortunately, it became even less clear what we were talking about.

Slurred language+forum format = misunderstanding with high probability.

 

GPT 4 turbo. Dialogues about the main thing.


 

Now about FF.