Better NN EA development - page 69

[Linuxser]

finimej:
I've double posted the code, can not edit my own code any more?

Fixed

[solaiman]

good, finimej.

[Orest]

finimej,

That's good you discovered the bug, but it's strange the EA gives 100% winners during the run within teaching period. It means it classifies trades correctly. Not trying to criticize you or anything, just a note on the side. I'm still interested in the fixed version though, please keep posting.

finimej:
I've double posted the code, can not edit my own code any more?

Here is the all the lectures notes

Pattern1.pdf

Pattern2.pdf

Pattern3.pdf

Pattern4.pdf

Pattern5.pdf

Pattern6.pdf

Pattern7.pdf

Pattern8.pdf

Pattern9.pdf

[finimej]

Orest above,

The EA trainer does not use the pnnclassificator at all. It backlooked and set the classficiation by the code from trainer itself. The pnn_system is using the classificator.

It's hard to debug somebodyelse elegent code, where it mixed global variable and local variable, and lots arrays.

I think better to write our own code, then I would like to have two files

pnn_buy_trained.dat

pnn_sell_trained.dat

then have two class array, one for buy and one for sell, to get the data to/from files. This would be easier to write the code and debug and also faster. just get no energy right now, somebody else take over the debug comez pnnclassification? The logic is written in the lecture notes pattern2.pdf and pattern3.pdf.

[finimej]

by the way, Does anyone would like to share his/her neuroshell 5.3 above + noxa cssa package with me? thanks.

[barnix]

Canon-ball problem using a Multi Layer Perceptron in MT4

with 4 input neuron and 6 hidden neuron:

#define IN 4

#define IN1 5 //IN+1

#define OUT 2

#define N_HU 6

#define N_HU1 7 //N_HU + 1

[barnix]

Canon-ball problem using a Multi Layer Perceptron in MT4

with 6 input neuron and 9 hidden neuron:

#define IN 6

#define IN1 7 //IN+1

#define OUT 2

#define N_HU 9

#define N_HU1 10 //N_HU + 1

[barnix]

Canon-ball problem using a Multi Layer Perceptron in MT4

with 8 input neuron and 12 hidden neuron:

#define IN 8

#define IN1 9 //IN+1

#define OUT 2

#define N_HU 12

[barnix]

Canon-ball problem using a Multi Layer Perceptron in MT4

with 18 input neuron and 27 hidden neuron:

#define IN 18 //2*9

#define IN1 19 //IN+1

#define OUT 2

#define N_HU 27 //3*9

#define N_HU1 28 //N_HU + 1

[shahinm]

To sum it up to this point

I have read this whole site plus some of the links - here is the results:

(2 systems) GOMEZ PNN & eucleadian matix have the most potential.

If Barnix gets paid for posts - then barnix is rich. No disrespect barnix. But how is a cannon ball so related to the financial markets