Artificial neural networks. - page 5
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
What matters is what we feed into the neural network inputs. It should be the relevant factors that influence the price.
There are three pillars on which the shaky foundation of time series networks rests. Or to be more precise - it kills the effort outright
Thecomplexity, repeatability and inconsistency of the training sample.
This is what you will encounter when looking for inputs.
The whole problem with forex in particular is that the data from the series (from the chart chart in simple terms) is repetitive and inconsistent.
This means that the golden rule of NS for the market comes into play: if we know what to input to NS, then NS is no longer needed.
sergeev:
...if we know what to feed into the NS, the NS is no longer needed.
And I was beginning to think I was the only one who thought so.
Neurons are mysterious to those who don't know them and therefore are attracted as a learning object. There is nothing wrong with that. Study, try, but don't exaggerate the power of networks as magic tools for revealing patterns in the market.
...if we know what to input to NS, then NS is no longer needed...
is not obvious.
If we know what to input to the NS, it does not mean that we can do without it (which also does not mean that the NS will help to make a decision based on the information input to it) and that there is a way to process this information sufficiently to make an unambiguous trading decision at all.
is not obvious.
it seems to me that if you start to solve the problem of repeatability and inconsistency, it becomes quite obvious
No, I'm not defending NS. But it's really not obvious. Well, if there is a better way of processing information that solves the same problems as NS, and it is reliably known, then yes.
But the problems of repeatability and inconsistency will not go anywhere, regardless of the way the information is processed.
Can you tell me what to input into the NS? =)
Can you tell me what to feed to the NS input? =)
No, we won't. Because if we do, then:
Networks are useful in the market for only two purposes:
1. Writing articles about them and getting royalties, and
2. If the same Expert Advisor running on the intersection of wizards advertised as one using neural networks and created by 2 mathematicians, 2 physicists, 2 neuroscientists and 2 economists, it will bring many more investors willing to pay much higher prices than if it is advertised as an Expert Advisor running on the intersection of wizards and created by one programmer.
Networks are useful in the market for only two purposes:
1. Writing articles about them and getting royalties, and
2. If the same Expert Advisor running on the crossover is advertised as one using neural networks and created by 2 mathematicians, 2 physicists, 2 neuroscientists and 2 economists, it will bring many more investors willing to pay much higher prices than if it is advertised as an Expert Advisor running on the crossover and created by one programmer.
Right. And chess programs will never learn to play above second rate.
I've heard that somewhere before.
--
Vladimir, I hope your madness is temporary, and I'd hate for it to last for thirty years (like Marvin Minsky's).
It's fun otherwise, yeah.
;)