neural network and inputs - page 11
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
The beginning of the topic is interesting and then it all devolves into a clarification of points of view.
Figures and proof of one's point of view ended on page 1-2.
Is it realistic to feed?
I tried indices, quotes, cross pairs (with and without correlation), FFT, FHT. The only thing I have tried is to feed the weather....
I have made forecasts too, in various ways, from "tomorrow" to "the strongest wave".
The errors of the trained grids are higher than the practical application.
Further build-up of "power", I think is a dead end.
my opinion, neural nets are a 'thing', but you need to work on them forever. That is, if you decide to use them, you should know: "there will be no free money". It amounts to a job for which you will get your reward.
------------------------------------------------------------------------
Serve up what MAY BE fish, purely logically. And you serve all sorts of standard turkeys, where there is no fish and can not be. I mean, look - for example, I'm sitting in the bathroom, and an idea comes to me. After thinking about it, I come to the conclusion that there might be fish in there... and then I go and try to catch it, and sometimes I use NS too (though rarely, just when I don't have the energy or time to deal with a regression whose shape I don't know from the beginning). But you're dancing from the opposite - a neuronet will do everything by itself. And this is not right.
IronBird:
- For example, I'm sitting in the bathroom and an idea comes to me...
...I jump out of the bathroom and yell, like Archimedes, "Eureka!!!"
;)))
It's like in the DMB about the jerboa. You don't see it, but it's there :-)
...I jump out of the bathroom and yell, like Archimedes, "Eureka!!!"
;)))
No. Then I smoke, close my eyes and start thinking about how to harness the idea better. It's good to think in the bathroom... The house doesn't bother me, the phone doesn't ring, the TV doesn't mutter something...
I agree. The network itself cannot do anything. I.e. it will not be possible to obtain a result purely by the architecture or power of the network. It is convenient to use it when you don't see obvious regularities, but they are there........
And then, if everything works, it would be nice to disassemble this neural network and see (and understand) why it is actually regressing there. That is, without understanding the meaning of the found effect it is elementary scary to trade.
The sense of the effect is simple - if there is a pattern or patterns between input and output data, the net will find them. If not, it is useless to look for them)))
Therefore, there is no need to dissect the net to find the meaning of the effect found )))
I am saying that if a grid has been found, it is a good idea for the creator to get a thorough grasp of it. And after that it will be possible to replace the grid with something more suitable - for processing exactly these regularities, this type of so-and-so. So as not to sell a black box
You say everything correctly, but there is a very big and thick BUT:
If there is no pattern in the data, but there are random coincidences in the analysed section, the NS will find them and give them out as a pattern.
If in some part of the trajectory of a ballistic missile with a nuclear warhead, its path coincided with the route of seasonal migration of cranes, then this does not mean that we can predict the entire route of the missile following a known migration path. But the NS will output the result as a revealed pattern.