Neural networks. Questions from the experts. - page 23

 
goga:

and all of this has to compensate each other. and neural networks have to check the signals of each block, look for patterns. and fully control and learn.

Is a 4 core processor enough for that?


Aftomatics on the verge of fantasy?

Anreal.

 
goga:
this is understandable, but it is not suitable for continuous comparison
why? so neural networks can do everything, even continuous comparisons, but the good old method cannot be switched on continuously ) and how is it continuous? with every tick?
 
goga: Why not? What are they for?
I wrote above about what they are for. In principle, you could also adapt a meat grinder to do this. But there's no point in going to the trouble....))))
 
alexeymosc:
why? so neural networks can do everything, even continuously compare, but the good old method cannot be switched on continuously ) and how is it continuous? with every tick?
For example, every 5 minutes. I just want to sleep sometimes, too.
 
In your question, it is much more important how you are going to teach the machine to distinguish between different market states.
 
LeoV:

No, it's an attempt to use non-linear transformations (neural networks) to find patterns in financial time series.

Is that what you mean?

Well, I mean finding patterns in financial timeseries plus finding patterns in block layout to find the best option.

 
goga: I am talking about searching for patterns in financial and time series, plus searching for patterns in block layout to find the best option.

You have to understand the complexity of the task you are posing. The human brain, with its huge number of neurons, cannot cope with such a task. What do you want from a poor neural network with only a few neurons? ))))
 
alexeymosc:
In your question, what is more important is how you are going to teach the machine to distinguish between different market states.
If you want to rephrase that, how do you teach an automaton to sense changes in market conditions?
 
LeoV:

You must understand the complexity of the task you are setting. The human brain, with its huge number of neurons, cannot cope with such a task. What do you want from a poor neural network with only a few neurons? ))))
I cannot agree with you there. the human brain can do it. the problem is that people do not devote enough time and attention to it and are designed for other tasks. but as i understand there are individuals who can only live on the informational plane
 

I think you need more than one network, depending on machine capacity, but they should all be interconnected even though they will be responsible for different processes