Using artificial intelligence at MTS - page 28
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
I've been thinking about this thought experiment:
...
I, too, cannot understand people who try to understand the phenomenon called 'mind' by studying the brain of living beings
The mind is just a technology...
Why study the "equipment" that implements the technology? It's like studying computer "hardware" while trying to understand how an archiver manages to "compress" files
Я тоже никак не могу понять людей, которые пытаются понять суть явления, имеющего название "разум", изучая мозг живых существ
The mind is just a technology...
I think it is more subtle and complicated than that. After all, animals, too, sometimes exhibit behaviour similar to that of a neural network (in particular, the ability to be trained). Humans differ from them in the presence of reflexivity: the ability to become aware of the motives of one's behaviour. Rather, my post demonstrated that using a neural network is not a panacea, whichever is used: an algorithm or an animal. And a rather direct analogy can be drawn between the two (the random decisions of a cockroach and the random wandering of a price chart).
I want to make some more remarks to the article under discussion, and for that I need to remember again about the cockroach.
How many electric shocks can the cockroach withstand before it either dies or gets sugar? (It is assumed, apparently, that by eating sugar, it regains its full strength and health.) The question is actually not idle: after all, an unkillable cockroach is equivalent to an "unkillable" (infinite) deposit. Of course, you can use a very large (almost infinite) demo deposit at the training stage. On the other hand, will a practically indestructible cockroach learn anything? He will pay attention only to "carrot", not to "stick", which already violates the analogy with forex, which does not forgive errors. On the other hand, if the cockroach dies before time, you will have to train a new one.
Yuri Reshetov, when testing your strategy, you should first of all demonstrate not equity (which may be positive or negative at random), but convergence of the linear filter coefficients to some so-called "true" values. Moreover, the convergence should be pronounced: the greater the length of the training sample, the smaller the distance between the experimental value and the true value.
Why can the equity be negative? You keep saying that the neural network can only determine the direction of price movement, but not its value. Imagine that for every 10 successful trades there is one unsuccessful one, which overrides all profits from successful ones. Then a profitable strategy turns into a losing one, despite the fact that formally there are more profitable trades than unsuccessful ones. It is necessary to set each trade the same ST and TP (not in the sense of ST=TP, but the same for each trade) - only then will the number of successful trades be decisive and it can be tested. I don't think you said a word about this in your article.
Finally, once the ST and TP are placed, you need to somehow prove that there will always be more "lucky" points on the chart on the same side of the plane than "unsuccessful" points. But I have no idea how this could be proved. Probably not at all (because it probably isn't).
... Even so, animals sometimes exhibit behaviour similar to that of a neural network...
That's understandable.
But, the very fact of existence of your post shows that there are people who, for some reason, instead of examining data for the presence of dependence in it with effective SPECIALIZED "tools", spend time and effort on fiddling with far from the best "equipment" (brains, neural networks).
Why? Why...? (not clear
)