Machine learning in trading: theory, models, practice and algo-trading - page 577

 
Yuriy Asaulenko:
7.0 is available on the web. Maybe it is already there?

Yes, I will take a look, of course.

Judging by whatUladzimir Izerski writes here,NS is a gold mine, we have to study it more carefully.

Alexander_K2:
If you mean the deals that are good, this is the most brilliant phrase :)))))
Uladzimir Izerski

There are already too many of them. It is necessary to choose the best of the best, and how to do that I do not know.

Time to go away from here...

 
Alexander_K2:

It's time to get out of here...


Me too.

 
Uladzimir Izerski:

Me too.


What kind of NS is it?

 
Maxim Dmitrievsky:

What kind of NS is it?

Why are you picking on the man? NS is like NS. They all look the same to me). The only question is what is in the inputs-outputs and how it is trained. But I don't think anyone can tell you that.)
 
Yuriy Asaulenko:
Why are you picking on the man? NS is like NS. For me, they are all the same.) The only question is what is in the inputs-outputs and how they are trained. But I don't think anyone can tell you that.)

Now, google it and write which one :)

 
Maxim Dmitrievsky:

Now, google it and write which one :)

By the way, Maxim, tell me please, how many inputs can RF realistically process in real time? I'm not very good at RF, just the basic principles, that's all.
 
Yuriy Asaulenko:
By the way, Maxim, tell me, how many inputs may RF really process in real time? I'm not very good at RF, just the basic principles and that's all.

over and very fast... but trained works slower than NS, because you have to go through all the trees. If on every tick to pick up the result will be significantly slower than with NS. If it's byte-by-tree, you won't notice it.

Non-interactive learning, i.e. instantaneous as compared to NS

 
Maxim Dmitrievsky:

over and very fast... but trained works slower than NS, because you have to go through all the trees. If on every tick to pick up the result will be significantly slower than with NS. If it's byte-by-tree, you won't notice it.

Non-interactive learning, i.e. instantaneous as compared to NS.

I see.

NNS with 15 inputs and 6 layers (60 neurons) - response time 0.005 sec. For NS I set the limit - maximum 15-20 inputs. If RF is slower, they are not an option anymore(.

 
Maxim Dmitrievsky:
I have already posted in my codebase the implementation of a linear spread indicator. If it does not work, I will do it too :)) But this one is much more tricky :)

Why would you post things that don't work? People will be wasting their time... or even their money.

 
Yuriy Asaulenko:

I see.

NS of 15 inputs and 6 layers (60 neurons) - response 0.005 s. For NS I set the limit - maximum 15-20 inputs. If RF is slower, they are no longer an option, it turns out.


Well, it depends on the implementation