How to form the input values for the NS correctly. - page 3
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
To StatBars something your file is not decompressed please check.
The topic has been abandoned for nothing. It's a very interesting one.
Apart from normalisation of input values, properly formed inputs can also include pattern selection or what exactly do you want to train the network to do?
If anyone has any ideas or experience, please share.
The topic has been abandoned for nothing. It's a very interesting one.
Apart from normalisation of input values, properly formed inputs can also include pattern selection or what exactly do you want to train the network to do?
If you have any ideas or experiences please share.
Yes, the topic should not be abandoned. Imho very much can teach or suggest to a novice neural networker :)
I think experience is very important in this business. I am therefore asking for the advice of the experienced.
I am currently studying the books:
Simon Haikin. I recommend for beginners the book NeuralNetworks by F. Wassermann Neurocomputer Science for the first acquaintance. Further, when everything is more or less clear in your head, you can systematize by D. Ivanov Financial Market Forecasting with Artificial Neural Networks.
All these authors write about the importance of input data.
In Simon's book I was also interested in the coupled gradient method. Can anyone share it, because everything in the book is very mathematical.
Roughly speaking I outline a plan of work with neuronet, or rather things to be paid attention to when developing it.
1. Preparing input data. (shifting averages, decorrelation, covariance equalization).
2. Correct outputs (ranges, extremes, directions)
3. The issue of retraining the network
4. Cross-checking
5. Network adaptation to new data
6. Conjugate gradient search optimization
7. Ability to use secular maps (or Kohonen and Grossberg layers?)
8. Committee of networks.
9. Recursive networks.
All this is to be studied and applied in my practice (plans are Napoleonic). And if all gurus have faced 1 and 2 points, then perhaps they can answer these two questions better than any theoretical book.
I'm currently researching-reading books:
Simon Haykin. Neural networks - he writes very well, but for beginners I would still recommend the book by F. Wasserman Neurocomputer Technology for the first acquaintance. Then, when everything becomes more or less clear in your head, you can systematize it with the book by D. Ivanov, Forecasting Financial Markets Using Artificial Neural Networks.
------------------------------------------------------
Alexey, if it is possible, give me the links to the mentioned books for free download.
If moderators allow, I'm posting the archives here (you'll have to delete them after reading them :))))
Thanks, uploaded it. Looked it up, seems to be missing.
Simon Heikin. Neural networks.
....
Just wrote this and it came up...
Copy part two. Will there be more?
Copy part three. Any more?
Copying part four. Any more?
Finished with Haikin, moving on....
Judging by the volume, it looks like we have a life to put down...
No, you won't have to lay down your life, just half of it... :)
All right, thanks, I copied Yezhov too. I'm going to try to get into these neural networks for real. In itself the idea has always pleased me.
But in respect to forex I was always doubtful, because "patterns" are bound to absolute values.