Machine learning in trading: theory, models, practice and algo-trading - page 606
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
That's why overestimating the number of neurons is also harmful. It won't be generalizing, it will be remembering along with the noise.
Well, either regularization/early stop... but you can't do that by yourself :))) and I feel that even the packages will be a pain...
here's an article I found... here's a summary of everything http://ai-news.ru/2016/05/tehnologii_fondovogo_rynka_10_zabluzhdenij_o_nejronnyh_setyah_578372.html
Theorists !!!
When will be the practice?
Not to repeat myself, I hope that Nikolai does not mind quoting. I completely agree with him.
Can't you cope with a simple approximator? ))
...
The NS does not retrieve any signs, signs are fed to the input. It either downsamples or memorizes all combinations (as the number of neurons increases).
I see, you extract the signs yourself and then give them to the input of the NS. If you already have signs, why do you need NS?
Apparently we read different books, and in different ways.)
Apparently we read different books, and in different ways.)
What is there to read, more neurons - more overtraining. Fewer neurons - worse approximation. Norm neurons - all normal, but it still does not work in the market in general due to non-stationarity.
I see, you extract the signs yourself, and then you feed them to the NS input. If you already have the signs, why do you need NS?
Apparently we read different books, and in different ways.)
Well, describe in more detail what and how you feed it, draw a schematic diagram... otherwise it's hard to guess
I did the classic method, as usual... I fed the garbage at the input and got garbage at the output.
Package learningCurve, R, Learning Curve.
Yeah, ok, thanks ) maybe I'll use it later
Well, describe in more detail what and how you feed, draw a schematic diagram... otherwise it's hard to guess
I did the classical way, as usual... I fed the trash at the input and got the trash at the output
(You are from the schemes.) (Heikin has all this. Do you want me to rewrite it?)
What do you mean by convolution or what? He has a lot of things, as much as a thousand pages :D
If you don't learn NS, it means you have convolution, i.e., data compression... Number of inputs and outputs in NS is about the same... There can only be this option
but MLP is a supervised method, and it's strange to use it as a convolution
You mean convolution or what? He has a lot of things, as much as 1000 pages :D
If you don't learn NS, it means you have convolution, i.e., data compression... Number of inputs and outputs in NS is about the same... There can only be this option.
We started with layers.)
What about what I have and how I learn, so I told you in detail, and even more than once.
We started with layers).
As for what I have and how I teach, I told you about it in detail, and even more than once.
15-20 inputs of whatever, + monte carlo added to it in whatever way... and the price is still being fed
that's not detailed :)
All in all, everything is clear with layers