I am testing the PNN code from Paco Hernández Gómez. The only thing that I did is after trained the pnn-trainer, I then optimized the limit and stop for last 2 days.

I then put the "PNN trainer" as expert assistant on EUR USD M1 and at the same time put the PNN system as EA on another EURUSD M1 chart.

Look at last 48 hours running results! It picked up exakt at the zigzag turnning points! However, it only make it to BUY operation. So if this is so nice can pick up the pattern for zigzag turning points, then only help is need is some help to get the right direction, such as AND if stochastic.. and RSI turns, then follow the direction.

Attached Images

pnn.gif (49.0 KB, 664 views)

Hi Finimej,

I am also testing the PNN but i keep getting the following error in the expert log

PNN EURUSD, M1: zero Divide

it indicates that the library file PNN.mq4 is having issues with a certain division but i can't seem to find out what or where . Any suggestion

Hello everybody,

I followed this thread from the beginning and I miss some information.

Can you explain me what is an input data for training of NN in the EA ??

What sort of optimalization is being done for input paramaters of NN, and NN itself, how often etc.

Thanks, Krzysztof

Hi, progressapama,
I have not changed anything in the code at all. Download and compile and works. On the FXDD demo account.
Backtest, used the option "open price on each bar" to training the pnn traininger, with selected week period. then use the EA to use the week after the selected period, the result is so: PNN can doing will direct for the first 2 days after the training data, and start to less it's edge after 2 dags good winning ratio, for the days in the week.

Real time demo forwarding test.
1) traing PNN trainer with option "every tick in the bar" use the 3 days data (included current day), it takes whole night to finish it.
2) then put PNN trainer as EA on EURUSD M1 chart, and put EA on another chart of EURUSD M1.

Forwarding test result is : it can realy pick up the trade exakt at the turning point with very precision entries. Howerver, it's so frustrated that the EA only do buying trading. (Although it can do sell trade in the backtest, so it is not the codes fault). I put it "only Short", then the EA is not doing anytrade at all. An EA can do SELL as well, is very much appreciated when we have a financial crise, isn't ? hasn't check out the code yet.

See forwarding test PNN system on demo account, has changed any code.
It has realy nice entries on all the right points and does not take any trading when the EURUSD is congesting. It is also stupid enough only do buy operations!

Attached Images

pnn.gif (20.6 KB, 429 views)

Barnix, maybe you can answer me those question ??

Krzysztof

The input values you can find in indicator source:
http://www.forex-tsd.com/attachments...tribind_2d.mq4

FANN library 2.0.0
Better NN EA development
http://leenissen.dk/fann/
Type: fann_type
fann_type is the type used for the weights, inputs and outputs of the neural network.

fann_type is defined as a:
float - if you include fann.h or floatfann.h
double - if you include doublefann.h
int - if you include fixedfann.h (please be aware that fixed point usage is
only to be used during execution, and not during training).
==================================
struct fann_train_data
{
enum fann_errno_enum errno_f;
FILE *error_log;
char *errstr;

unsigned int num_data;
unsigned int num_input;
unsigned int num_output;
fann_type **input;
fann_type **output;
};

Read fann_train_data from file:
xor.data

4 2 1
-1 -1
-1
-1 1
1
1 -1
1
1 1
-1

/*
* INTERNAL FUNCTION Reads training data from a file descriptor.
*/
struct fann_train_data *fann_read_train_from_fd(FILE * file, const char *filename)
.....
if(fscanf(file, "%u %u %u\n", &num_data, &num_input, &num_output) != 3)

num_data=4
num_input=2
num_output=1

4 2 1 //num_data, num_input, num_output
-1 -1 //input
-1 //output
-1 1 //input
1 //output
1 -1 //input
1 //output
1 1 //input
-1 //output

Error structure:
/* Struct: struct fann_error

Structure used to store error-related information, both
<struct fann> and <struct fann_train_data> can be casted to this type.

See also:
<fann_set_error_log>, <fann_get_errno>
*/
struct fann_error
{
enum fann_errno_enum errno_f;
FILE *error_log;
char *errstr;
};

===============================
/* INTERNAL FUNCTION
Initialize an error data strcuture
*/
void fann_init_error_data(struct fann_error *errdat)
{
errdat->errstr = NULL;
errdat->errno_f = FANN_E_NO_ERROR;
errdat->error_log = fann_default_error_log;
}