NeuroBook/Scripts/batch_norm/perceptron_test_norm.mq5

<EFBFBD><EFBFBD>//+------------------------------------------------------------------+

//|                                              perceptron_test.mq5 |

//|                                  Copyright 2021, MetaQuotes Ltd. |

//|                                             https://www.mql5.com |

//+------------------------------------------------------------------+

#property copyright "Copyright 2021, MetaQuotes Ltd."

#property link      "https://www.mql5.com"

#property version   "1.00"

#property script_show_inputs

//+------------------------------------------------------------------+

//| External parameters for script operation                         |

//+------------------------------------------------------------------+

input string   StudyFileName  = "study_data.csv";  // Name of file with the training dataset

input string   OutputFileName = "loss_study_bn.csv";  // File name to write error dynamics

input int      BarsToLine     = 40;                // Number of historical bars in one pattern

input int      NeuronsToBar   = 4;                 // Number of input layer neurons per 1 bar

input bool     UseOpenCL      = false;             // Use of OpenCL

input int      BatchSize      = 10000;             // Batch size to update the weight matrix          

input double   LearningRate   = 3e-5;              // Learning rate

input int      HiddenLayers   =  1;                // Number of hidden layers

input int      HiddenLayer    =  40;               // Number of neurons in one hidden layer

input int      Epochs         =  1000;             // Number of wight matrix update iterations

//+------------------------------------------------------------------+

//| Connect the neural network library                               |

//+------------------------------------------------------------------+

#include "..\..\Include\realization\neuronnet.mqh"

//+------------------------------------------------------------------+

//| Script program start                                             |

//+------------------------------------------------------------------+

void OnStart(void)

  {

   VECTOR loss_history;

//--- prepare vector for storing the history of network errors

   if(!loss_history.Resize(0, Epochs))

     {

      Print("Not enough memory for loss history");

      return;

     }

   CNet net;

//--- 1. Network initialization

   if(!NetworkInitialize(net))

      return;

//--- 2. Loading training set data

   CArrayObj data;

   CArrayObj result;

   if(!LoadTrainingData(StudyFileName, data, result))

      return;

//--- 3. Training the network

   if(!NetworkFit(net, data, result, loss_history))

      return;

//--- 4. Save the history of network errors

   SaveLossHistory(OutputFileName, loss_history);

//--- 5. Save the resulting network

   net.Save("Study.net");

   Print("Done");

  }

//+------------------------------------------------------------------+

//| Load training data                                               |

//+------------------------------------------------------------------+

bool LoadTrainingData(string path, CArrayObj &data, CArrayObj &result)

  {

   CBufferType *pattern;

   CBufferType *target;

//--- open the file with the training dataset

   int handle = FileOpen(path, FILE_READ | FILE_CSV | FILE_ANSI | FILE_SHARE_READ, ",", CP_UTF8);

   if(handle == INVALID_HANDLE)

     {

      PrintFormat("Error of open study data file: %d", GetLastError());

      return false;

     }

//--- display the progress of loading training data in the chart comment

   uint next_comment_time = 0;

   enum

     {

      OutputTimeout = 250 // no more than 1 time every 250 milliseconds

     };

//--- organize loop to load training dataset

   while(!FileIsEnding(handle) && !IsStopped())

     {

      if(!(pattern = new CBufferType()))

        {

         PrintFormat("Error creating Pattern data array: %d", GetLastError());

         return false;

        }

      if(!pattern.BufferInit(1, NeuronsToBar * BarsToLine))

         return false;

      if(!(target = new CBufferType()))

        {

         PrintFormat("Error creating Pattern Target array: %d", GetLastError());

         return false;

        }

      if(!target.BufferInit(1, 2))

         return false;

      for(int i = 0; i < NeuronsToBar * BarsToLine; i++)

         pattern.m_mMatrix[0, i] = (TYPE)FileReadNumber(handle);

      for(int i = 0; i < 2; i++)

         target.m_mMatrix[0, i] = (TYPE)FileReadNumber(handle);

      if(!data.Add(pattern))

        {

         PrintFormat("Error adding training data to array: %d", GetLastError());

         return false;

        }

      if(!result.Add(target))

        {

         PrintFormat("Error adding training data to array: %d", GetLastError());

         return false;

        }

      //--- show loading progress in the chart comment (no more than 1 time every 250 milliseconds)

      if(next_comment_time < GetTickCount())

        {

         Comment(StringFormat("Patterns loaded: %d", data.Total()));

         next_comment_time = GetTickCount() + OutputTimeout;

        }

     }

   FileClose(handle);

   return(true);

  }

//+------------------------------------------------------------------+

//| Initializing the network architecture                            |

//+------------------------------------------------------------------+

bool CreateLayersDesc(CArrayObj &layers)

  {

   CLayerDescription *descr;

//--- create input data layer

   if(!(descr = new CLayerDescription()))

     {

      PrintFormat("Error creating CLayerDescription: %d", GetLastError());

      return false;

     }

   descr.type         = defNeuronBase;

   int prev =

      descr.count     = NeuronsToBar * BarsToLine;

   descr.window       = 0;

   descr.activation   = AF_NONE;

   descr.optimization = None;

   if(!layers.Add(descr))

     {

      PrintFormat("Error adding layer: %d", GetLastError());

      delete descr;