Better NN EA - Page 8

mikkom · #71 (**permalink**) 12-05-2007, 01:26 PM

Quote:

Originally Posted by Craig

All TA is curve fitting, future predictions based on patterns in the past.
NNs are just another way of looking for these patterns.

Exactly my point. The problem with nn's is that to train them you have to understand the theory behind them very well.

ps. I would take a look at genetic programming if you are into this stuff, it's in some ways much cleaner than nn's and the information and algorithms gained with genetic algorithms can be used as part of your normal algorithms as well.

For more info see
Genetic programming - Wikipedia, the free encyclopedia

or read couple of Kozas books.

antone · #72 (**permalink**) 12-05-2007, 01:58 PM

Can anyone recommend to me a good NN tutorial book for newbies? i wanna learn it..

i have been interested with NN's way before better's EA..

not just for EA,, but also for sports picking..

mikkom · #73 (**permalink**) 12-05-2007, 02:41 PM

Bishops book is quite good

Amazon.com: Neural Networks for Pattern Recognition: Books: Christopher M. Bishop

Strangguy · #74 (**permalink**) 12-06-2007, 12:22 AM

Quote:

Originally Posted by project1972

The NN logic is not linear, Better EA sometimes trade Counter-Trend and Sometimes Trend-Following.

You can't back-engineering a non linear logic, most of the time not even the author know what are rules learned by a NN logic, it's a black-box for everyone including his own creator.

You can check the past trades, but you don't have a clue what rules will trigger the next.

Dear Project1972,
I have logged into this account with MetaTrader but I can only see his current trade on the chart. How did you get the start and ending icons of previous trades to show on the chart you posted?

Thanks,
StrangeGuy

primajaya · #75 (**permalink**) 12-06-2007, 03:47 AM

Quote:

Originally Posted by barnix

Naive Bayesian Classifier (=> PNN used by Better)
The probabilistic neural network is a direct continuation of the work on Bayes classifiers.

Thank's Barnix for your nice post

Pj

tdion · #76 (**permalink**) 12-07-2007, 02:36 PM

This is because there are three NNs trading within Better.

It is impossible to group what NN is posting what signals.

Quote:

Originally Posted by project1972

The NN logic is not linear, Better EA sometimes trade Counter-Trend and Sometimes Trend-Following.

You can't back-engineering a non linear logic, most of the time not even the author know what are rules learned by a NN logic, it's a black-box for everyone including his own creator.

You can check the past trades, but you don't have a clue what rules will trigger the next.

LesioS · #77 (**permalink**) 12-10-2007, 09:03 AM

Quote:

Originally Posted by gbolla

same ideas of writing EA like Better's EA?
how is it possible write a complex neural nets in mql language? And about a backpropagation error algorithm ? or he training the net on C++ and then re-write in mql with only the weights?? how many hidden layer?

bolla

Better doesn't use backpropagation nets... I suppose that it would be to difficult and resource consuming to use it during automatic trade. Better uses probabilistic networks which are more simplier to implement and retrain during real trade.

barnix · #78 (**permalink**) 12-10-2007, 09:41 AM

311: Neural Networks
3. Implementing PNN

// C is the number of classes, N is the number of examples, Nk are from class k
// d is the dimensionality of the training examples, sigma is the smoothing factor
// test_example[d] is the example to be classified
// Examples[N][d] are the training examples

int PNN(int C, int N, int d, float sigma, float test_example[d], float Examples[N][d])
{
int classify = -1;
float largest = 0;
float sum[ C ];
// The OUTPUT layer which computes the pdf for each class C
for ( int k=1; k<=C; k++ )
{
sum[ k ] = 0;
// The SUMMATION layer which accumulates the pdf
// for each example from the particular class k
for ( int i=0; i<Nk; i++ )
{
float product = 0;
// The PATTERN layer that multiplies the test example by the weights
for ( int j=0; j<d; j++ )
product += test_example[j] * Examples[i][j];

product = ( product – 1 ) / ( sigma * sigma );
product = exp( product );
sum[ k ] += product;
}
sum[ k ] /= Nk;
}
for ( int k=1; k<=C; k++ )
if ( sum[ k ] > largest )
{
largest = sum[ k ];
classify = k;
}
return classify;
}

ak97052d · #79 (**permalink**) 12-11-2007, 09:01 PM

PHP Code:


			
//+------------------------------------------------------------------+
//|                                                          PNN.mq4 |
//|                                          Copyright © 2007, Klot. |
//|                                            http://www.fxreal.ru/ |
//+------------------------------------------------------------------+
#property copyright "Copyright © 2007, Klot."
#property link      "http://www.fxreal.ru/"
#property library

//+------------------------------------------------------------------+
//| C - is the number of classes, N - is the number of examples,     |
//| Nk - are from class k                                            |
//| d - is the dimensionality of the training examples,              |
//| sigma - is the smoothing factor                                  |
//| test_example[d] - is the example to be classified                |
//| Examples[N][d] - are the training examples                       |
//+------------------------------------------------------------------+
int PNN(int C, int N, int Nk, int d, double sigma, double test_example[d], double Examples[N][d]) 
   { 
   int classify = -1; 
   double largest = 0; 
   double sum[];
   ArrayResize(sum,C);
   // The OUTPUT layer which computes the probability density function for each class C 
   for ( int k=0; k<C; k++ ) 
      {
      sum[ k ] = 0; 
      // The SUMMATION layer which accumulates the probability density function  
      // for each example from the particular class k 
      for ( int i=0; i<Nk; i++ ) 
         {
         double product = 0; 
         // The PATTERN layer that multiplies the test example by the weights 
         for ( int j=0; j<d; j++ )  product += test_example[j] * Examples[i][j];
         //-----
         product = (product-1)/(sigma*sigma); 
         product = MathExp(product); 
         sum[k] += product;
         } 
      sum[k]/=Nk;
      }
   //----
   for ( k=0; k<C; k++ )
   if ( sum[ k ] > largest ) 
      {
      largest = sum[ k ]; 
      classify = k;
      }
   //----
   return (classify);
   } 
//+------------------------------------------------------------------+

From this SITE

gbolla · #80 (**permalink**) 12-13-2007, 02:01 PM

thanks...but How do you use this piece of code in EA?
and about parameters?? what's mean?
thanks

Bolla

LinkBacks (?) LinkBack to this Thread: http://www.forex-tsd.com/expert-advisors-metatrader-4/10848-better-nn-ea.html
Posted By	For	Type	Date
Harmony of Life - Neural Network (The Artificial Intelegence)	Post #8	Refback	03-14-2008 09:30 AM
Harmony of Life - Neural Network (The Artificial Intelegence)	Post #8	Refback	12-20-2007 01:58 AM
Harmony of Life - Aditya's Blog	Post #8	Refback	12-05-2007 11:06 AM
Forex MetaTrader Expert Advisors -- Select From >40 of the Best Expert Advisors EA for MT4 on Squidoo	This thread	Refback	11-24-2007 08:56 AM
Forex MetaTrader Expert Advisors -- Select From >40 of the Best Expert Advisors EA for MT4 on Squidoo	This thread	Refback	11-23-2007 04:54 PM

Register in Forex TSD!
Trading Systems Leaders in this forum (automated trading systems) are winning more than 3000 pips in a month (30000$ investing one lot every time). Click here to register and get more information