Better NN EA development - page 30
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
From previous post:
noc_withoutnn.htm@Dimicr - I see a test using Cortex weights - are you using the 'original' weights, or have you managed to train a network on current price-data???
I cannot get Cortex to work properly. Training the NOC indicator via scripts is not a problem, it's just the weights generated are useless...
I use "original" weights from Cortex site.
Yes, this version of Cortex don't work properly. May be have bug in backpropagation dll - don't know, but generated weights are completely wrong.
Looking for Code to Generate Correct Weights
In Cortex have something wrong. I train with my own backpropagation code this indicator with data from 2007 to today. May be have forecast for one bar.
Hi Dimicr,
I run into the same problem. The weights in Cortex are generated by some other Code and not the Cortex code. I have been looking for the Weights generation Code for Months. Only the weights generation code so that i can generate weights for other Pairs like GBPUSD, USDJPY, USDCHF, etc . i also have my own BACKPROP code that i use to test the EA using the indicator. The Weights for EURUSD on the cortex website are Fine. But if you want to generate Code for other Pairs it is totally a mess.
Please reply with attachment of Weights Generation Code if possible . Thank you very much or email me . we can work together . I really need to generate the Weights for other currencies.
Thank you
I use my code and other tools for weghts generation. Here you can find open source backpropagation code basegroup
btw all that Cortex do is prediction for one bar of NOC indicator with NN. If one system isn't profitable with NOC - prediction is useless. 2 progressapama send me good results with attached system on other Pairs and i'l make NN NOC for them.
@Dimicr - thanks, I appreciate the response. One other question, what is your opinion of FANN?
One last note regarding Cortex - though the weights-list it provides seem well messed up, it seems to be 'back-propagating' OK, as the pic below shows (indicator & predicted value superimposed). As it doesn't allow access to the correct weights though, it's worthless!...
cortex problem
I have purchased all 3 cortex packages, and just contacted the author of the software to let him know about the weight problem will let you know what I hear back. Cheers
I have purchased all 3 cortex packages, and just contacted the author of the software to let him know about the weight problem will let you know what I hear back. Cheers
I looked in vain for a contact email address - could you post it? As a paid-up owner, it can only help if as many of us as possible complain...