ObjSim/sensorycancel/sensorycancel.cpp

#include <iostream.h>
#include <math.h>
#include <stdio.h>
#include <stdlib.h> 

#include <libcsim.h>
#include <anyoption.h>
#include <valarray>

#include <time.h>


////////////////////

const    float                       connectivity = 0.1;


int main(int argc, char** argv)
{

     clock_t cpu_start, cpu_end;
     double cpu_time_used;

////////////////Begin//ProcessCommandLineArguments//////////////
     cpu_start = clock();

// 'a' = dummy; do not use -a !! (setOption without second para doesn't work!!)

     AnyOptionWrapper *myAnyWrap = new AnyOptionWrapper(argc, argv,"settings_sensorycancel.cfg");
     float NoiseSigma; myAnyWrap->setOption( "NoiseSigma", 'n', &NoiseSigma, 0);
     float InputNoiseSigma; myAnyWrap->setOption( "InputNoiseSigma", 'n', &InputNoiseSigma, 0);
     float InhNoiseSigma; myAnyWrap->setOption( "InhNoiseSigma", 'n', &InhNoiseSigma, 0);
     float istrength; myAnyWrap->setOption( "InputStrength", 'i', &istrength, 1);
     float InitialWeights; myAnyWrap->setOption( "InitialWeights", 'w', &InitialWeights, 0.01);
     float InitialFWInhWeights; myAnyWrap->setOption( "InitialFWInhWeights", 'w', &InitialFWInhWeights, 0.02);

     float InputSat; myAnyWrap->setOption( "InputSaturation", 'S', &InputSat, 1.0);
     float ris;     myAnyWrap->setOption(  "RandomInputStrength", 'r', &ris, 10.0);
     float WeightSum;   myAnyWrap->setOption( "WeightSum", 'u', &WeightSum, 1);
     bool TestSim; myAnyWrap->setFlag ("Test", 'T', &TestSim);
     bool LoadWeights; myAnyWrap->setFlag ( "LoadWeights", 'L', &LoadWeights);
     bool LoadITWeights; myAnyWrap->setFlag ( "LoadITWeights", 'a', &LoadITWeights);
     bool LoadForwardWeights; myAnyWrap->setFlag ( "LoadForwardWeights", 'a', &LoadForwardWeights);
     float LtpInc;     myAnyWrap->setOption( "LtpInc", 'c', &LtpInc, 0.7);
     int MaxTrials; myAnyWrap->setOption("MaxTrials", 'a', &MaxTrials, 80);
     int probesteps=0; myAnyWrap->setOption("ProbeSteps", 'a', &probesteps, 4);
//     int learnsteps= 150;
     int learnsteps= 20; myAnyWrap->setOption("LearnSteps", 'a', &learnsteps, 20);

//      float TauRec; myAnyWrap->setOption("TauRec", 'a', &TauRec, 50);
//      float U_se; myAnyWrap->setOption("U_se", 'a', &U_se, 0.05);

     float BInpInhWeight;  myAnyWrap->setOption("BInpInhWeight", 'a', &BInpInhWeight, 0.03);   
     float MaxFWInhWeights;  myAnyWrap->setOption("MaxFWInhWeights", 'a', &MaxFWInhWeights, 0.1);   
     float FwdInhLinkingWeight; myAnyWrap->setOption("FwdInhLinkingWeight", 'a', &FwdInhLinkingWeight, 0.1);   

     float ExInhWeight;   myAnyWrap->setOption("ExInhWeight", 'a', &ExInhWeight, 0.1);   
     float InhExWeight; myAnyWrap->setOption("InhExWeight", 'a', &InhExWeight, 0.2);   

     float ITExInhWeight;   myAnyWrap->setOption("ITExInhWeight", 'a', &ITExInhWeight, 0.1);   
     float ITInhExWeight; myAnyWrap->setOption("ITInhExWeight", 'a', &ITInhExWeight, 0.2);   
     float ITExExWeight;    myAnyWrap->setOption("ITExExWeight", 'a', &ITExExWeight, 0.1);   

     float ITExCInhWeight;   myAnyWrap->setOption("ITExCInhWeight", 'a', &ITExCInhWeight, 0.001);   
     float ITCInhExWeight;   myAnyWrap->setOption("ITCInhExWeight", 'a', &ITCInhExWeight, 0.005);   
     float ITFBInhWeight; myAnyWrap->setOption("ITFBInhWeight", 'a', &ITFBInhWeight, 0.005);   
     float FBInhWeight; myAnyWrap->setOption("FBInhWeight", 'a', &FBInhWeight, 0.005);   
     float ITSelfInhibitionWeight; myAnyWrap->setOption("ITSelfInhibitionWeight", 'a', &ITSelfInhibitionWeight, 0.01);   

     float LearnRate; myAnyWrap->setOption("LearnRate", 'a', &LearnRate, 0.0001);
     float InhLearnRate; myAnyWrap->setOption("InhLearnRate", 'a', &InhLearnRate, 0.00001);
     float InhLearnBaseLine; myAnyWrap->setOption("InhLearnBaseLine", 'a', &InhLearnBaseLine, 1);
     float FBInhLearnRate; myAnyWrap->setOption("FBInhLearnRate", 'a', &FBInhLearnRate, 0.001);
     float Cdep; myAnyWrap->setOption("Cdep", 'a', &Cdep, 0.00001);
     float FBInhMaxWeight;myAnyWrap->setOption("FBInhMaxWeight", 'a', &FBInhMaxWeight, 1.0);


     float RossumCp; myAnyWrap->setOption("RossumCp", 'a', &RossumCp, 0.000001);
     float ITNormFrequency; myAnyWrap->setOption("ITNormFrequency", 'a', &ITNormFrequency, 30);


     bool RandomInputOrder; myAnyWrap->setFlag ("RandomInputOrder", 'a', &RandomInputOrder);

     float StimDur;  myAnyWrap->setOption("StimDur", 'a', &StimDur, 30);
     float IsiDur;  myAnyWrap->setOption("IsiDur", 'a', &IsiDur, 30);
     float TestStimDur;  myAnyWrap->setOption("TestStimDur", 'a', &TestStimDur, 100);
     bool Teach; myAnyWrap->setFlag ("Teach", 'a', &Teach);
     bool XContinuous; myAnyWrap->setFlag ("XContinuous", 'a', &XContinuous);
     float ExExRange; myAnyWrap->setOption("ExExRange", 'a', &ExExRange, 0.02);
     float ExInhRange; myAnyWrap->setOption("ExInhRange", 'a', &ExInhRange, 0.02);
     float InhExRange; myAnyWrap->setOption("InhExRange", 'a', &InhExRange, 0.08);
     float ExExMinDelay; myAnyWrap->setOption("ExExMinDelay", 'a', &ExExMinDelay, 3);
     float ExExMaxDelay; myAnyWrap->setOption("ExExMaxDelay", 'a', &ExExMaxDelay, 6);
     std::string DataDirectory; myAnyWrap->setOption("DataDirectory", 'a', &DataDirectory, "/home/frank/data/");


     float SelfInhMaxWeight; myAnyWrap->setOption("SelfInhMaxWeight", 'a', &SelfInhMaxWeight, 6);
     float SelfInhMinWeight; myAnyWrap->setOption("SelfInhMinWeight", 'a', &SelfInhMinWeight, 1);
     float SelfInhLearnTauDec; myAnyWrap->setOption("SelfInhLearnTauDec", 'a', &SelfInhLearnTauDec, 20);
     float SelfInhBaseLine; myAnyWrap->setOption("SelfInhBaseLine", 'a', &SelfInhBaseLine, 0.1);
     float SelfLearnRate; myAnyWrap->setOption("SelfLearnRate", 'a', &SelfLearnRate, 0.0001);
     float SelfInhWeightSum; myAnyWrap->setOption("SelfInhWeightSum", 'a', &SelfInhWeightSum, 5);

     myAnyWrap->process();
     delete myAnyWrap;
////////////////End//ProcessCommandLineArguments//////////////

////////////////Begin//InitializeGlobalSimLoop
     SimLoop* MainSimLoop = InitLibCSim(); 
//     MainSimLoop->SetDataDirectory("/home/frank/data/sim/csim/v4learn/");
     MainSimLoop->SetDataDirectory(DataDirectory.c_str());
     float dt=MainSimLoop->GetDeltaT(); // milli sec
     cout << "int main(): dt = " << dt << "\n";
////////////////End//InitializeGlobalSimLoop
     
     bool Learn=true;
     if ((LearnRate == 0) || TestSim) {
	  Learn=false;
	  cout << "No Learning\n";
     }

     int		i, j, k, sec, t;


//      lif2layer *ca3layer;
//      ca3layer=new lif2layer(N, dt, 20, 0.0033, 1, 4,4,ffi,fbi,2, 0.004, 10);

     if (TestSim) LoadWeights=true;
     
     if (LoadWeights)
     {
	  LoadITWeights=true;
	  LoadForwardWeights=true;
     }

////////////////End//ProcessCommandLineArguments//////////////

//////////////////////////////////////////////
     
     

//     ObjMovie MyMovie("/home/frank/prog/crystal/movies/crystal_all_dirchange.idlmov");
//     ObjMovie MyMovie("/home/frank/prog/crystal/movies/tetraeder02.idlmov");
//     ObjMovie MyMovie("/home/frank/prog/crystal/movies/verticalbar.idlmov");
//     ObjMovie MyMovie("/home/frank/prog/crystal/movies/horizontalbar.idlmov");
//     ObjMovie MyMovie("/home/frank/prog/crystal/movies/twobar_o3.idlmov");
//     ObjMovie MyMovie("/home/frank/prog/crystal/movies/twooris.idlmov");
//     ObjMovie MyMovie("/home/frank/prog/crystal/movies/16oris.idlmov");
//     ObjMovie MyMovie("/home/frank/prog/crystal/movies/16bars.idlmov");
//     ObjMovie MyMovie("/home/frank/prog/crystal/movies/winkel.idlmov");
//     ObjMovie MyMovie("/home/frank/prog/crystal/movies/winkelbars256.idlmov");
//     ObjMovie MyMovie("/home/frank/prog/crystal/movies/winkelbars64.idlmov");
     ObjMovie MyMovie("/home/frank/prog/crystal/movies/winkelbars2.idlmov");


     // generate InputLayers
     typedef izhlayer TInputLayer;
     TInputLayer* TmpInputLayer;
     vector<TInputLayer*> InputLayerList;

//////// Setup Input Layer
     int InputNx, InputNy;
     InputNx=InputNy=20;
     izh2layer* BlobInputLayer = new izh2layer(InputNx*InputNy, IzhParaIntegrator, InputSat);
     BlobInputLayer->SetName("BInpLayer");
     BlobInputLayer->SetupPositions(InputNx, InputNy, true);
     BlobInputLayer->SetRandomInputStrength(0);
     BlobInputLayer->SetNoiseSigma(InputNoiseSigma);
     BlobInputLayer->SetTauInh(7.0);
     MainSimLoop->AddSimElement(BlobInputLayer);



//////// Setup IT Module
     int ITNx, ITNy;
     ITNx = ITNy = 8;
//     ITNx = 1; ITNy = 2;
     IzhParas IzhLowDIntegrator   (0.02, -0.1, -55, 2, 108, 4.1, 0.0);
//     izh3layer* ITLayer = new izh3layer(ITNx*ITNy, IzhLowDIntegrator,InputSat);
     izh3layer* ITLayer = new izh3layer(ITNx*ITNy, IzhParaIntegrator,InputSat);
     ITLayer->SetName("ITLayer");
     ITLayer->SetupPositions(ITNx,ITNy,true);
     ITLayer->SetRandomInputStrength(0);
     ITLayer->SetNoiseSigma(NoiseSigma);
     ITLayer->SetTauInh(5.0);
     ITLayer->StartBinRec(2);
     MainSimLoop->AddSimElement(ITLayer);

     int ITInhNx, ITInhNy;
     ITInhNx = ITInhNy = ITNx/2;     
//     izh3layer* ITInhLayer = new izh3layer(ITInhNx*ITInhNy, IzhParaFastSpiking,InputSat);
     izh3layer* ITInhLayer = 0;
     if ((ITInhExWeight > 0) || (ITExInhWeight > 0)) {
	  ITInhLayer = new izh3layer(ITInhNx*ITInhNy, IzhParaClass2,InputSat);
	  ITInhLayer->SetName("ITInhLayer");
	  ITInhLayer->SetupPositions(ITInhNx,ITInhNy,true);
	  ITInhLayer->SetRandomInputStrength(0);
	  ITInhLayer->SetNoiseSigma(InhNoiseSigma);
	  ITInhLayer->SetTauEx(15);
	  MainSimLoop->AddSimElement(ITInhLayer);
     }

     int ITCommonInhNx, ITCommonInhNy;
     ITCommonInhNx = ITCommonInhNy = 2;     
     izh3layer* ITCommonInhLayer = new izh3layer(ITCommonInhNx*ITCommonInhNy, IzhParaFastSpiking,InputSat);
//    izh3layer* ITCommonInhLayer = new izh3layer(ITCommonInhNx*ITCommonInhNy, IzhParaClass2,InputSat);
     ITCommonInhLayer->SetName("ITCommonInhLayer");
     ITCommonInhLayer->SetupPositions(ITCommonInhNx,ITCommonInhNy,true);
     ITCommonInhLayer->SetRandomInputStrength(0);
     ITCommonInhLayer->SetNoiseSigma(0);
     ITCommonInhLayer->SetTauEx(10);
     MainSimLoop->AddSimElement(ITCommonInhLayer);



//////////connections

     FacilitativeConnection *ConITSelfInh=0;
     if (ITSelfInhibitionWeight > 0) {
// 	  connection *ConITSelfInh = new connection(ITLayer, ITLayer, 1, true);  // inputNumber=0   
	  ConITSelfInh = new FacilitativeConnection(ITLayer, ITLayer, 1, 60, 0.005, 50, 0.01);  // inputNumber=0   
	  ConITSelfInh->SetNonSelf(true);

	  ConITSelfInh->SetName("conIT_SelfInh");
	  if (LoadITWeights) ConITSelfInh->Load(); 
	  else {
	       ConITSelfInh->ConnectFull(ITSelfInhibitionWeight, 1,0);
//	       ConITSelfInh->ConnectRandom(0.4, ITSelfInhibitionWeight, 1,0);
// 	       DistanceProfile* profile = new CircleDistanceProfile(1000, 0.2);
// 	       ConITSelfInh->ConnectProfile(profile, ITSelfInhibitionWeight, 1,0, true);	       

//	       ConITSelfInh->ConnectRandom(0.2, ITSelfInhibitionWeight, 1,0);
	  }
	  ConITSelfInh->StartBinRec(1);
	  
	  if (Learn && (SelfLearnRate >0)) {
	       LearnHebbLP2* LearnSelfITInh = new LearnHebbLP2(
		    ConITSelfInh, SelfInhMaxWeight, SelfInhLearnTauDec, 
		    SelfInhBaseLine, SelfLearnRate);
	       LearnSelfITInh->SetMinWeight(SelfInhMinWeight);
	       ConITSelfInh->SetLearnObj(LearnSelfITInh);
	  } 
     
// 	  ConstSumNormalize* SelfInhNorm=0;
// 	  if (Learn) {
// 	       SelfInhNorm = new ConstSumNormalize(SelfInhWeightSum, false);     
// 	       SelfInhNorm->AddConnection(ConITSelfInh);
// //	       SelfInhNorm->CalcInitWeightSum();
// 	       SelfInhNorm->NormalizeAll();
// 	       MainSimLoop->AddSimElement(SelfInhNorm);
// 	  }

	  ConITSelfInh->Save();
	  ConITSelfInh->SetAutoSave(false);
	  MainSimLoop->AddSimElement(ConITSelfInh);
     }


     if (ITExCInhWeight > 0) {
	  connection *ConITExCInh = new connection(ITLayer, ITCommonInhLayer, 0);  // inputNumber=0   
	  ConITExCInh->SetName("conIT_ExCInh");
	  if (LoadITWeights) ConITExCInh->Load(); 
	  else {
	       ConITExCInh->ConnectFull(ITExCInhWeight, 1,0);
	  }
	  ConITExCInh->Save();
	  MainSimLoop->AddSimElement(ConITExCInh);
     }


//      if (ITCInhExWeight > 0) {
// 	  connection *ConITCInhEx = new connection(ITCommonInhLayer, ITLayer, 1);    
// 	  ConITCInhEx->SetName("conIT_CInhEx");
// 	  if (LoadITWeights) ConITCInhEx->Load(); 
// 	  else {
// 	       ConITCInhEx->ConnectFull(ITCInhExWeight, 1,0);
// 	  }
// 	  ConITCInhEx->Save();
// 	  MainSimLoop->AddSimElement(ConITCInhEx);
//      }

     float InhLearnTauDec=30;
     float InhLearnTauPre=30;
     float InhLearnTauPost=100;
     float FBInhMaxWeights=3;
     connection *ConITFbInh =0;
     if (ITFBInhWeight > 0) {
	  ConITFbInh = new connection(ITCommonInhLayer,  BlobInputLayer, 1);    
	  ConITFbInh->SetName("conIT_FBInh");
	  if (LoadITWeights) ConITFbInh->Load(); 
	  else {
	       ConITFbInh->ConnectFull(ITFBInhWeight, 1,0);
	  }

 	  if (Learn && (FBInhLearnRate>0)) {
	       ConITFbInh->SetLearnObj(
  	       new LearnFBInh(ConITFbInh, 
  			      FBInhMaxWeight, InhLearnTauPre,InhLearnTauPost, FBInhLearnRate, Cdep));
// 		    new LearnHebbLP2(ConITFbInh, 
// 				     FBInhMaxWeights, InhLearnTauDec, 0.1, LearnRate));
// //	  new LearnRossum1(ConITFBInh, RossumCp, 0.003, 0.0002));
	       
	  }

	  ConITFbInh->Save();
	  ConITFbInh->SetAutoSave(false);
	  MainSimLoop->AddSimElement(ConITFbInh);
     }


     connection *ConInpIT = new connection(BlobInputLayer, ITLayer, 0, false);
//     FacilitativeConnection *ConInpIT = new FacilitativeConnection(BlobInputLayer, ITLayer, 0, 300, 0.003, 50, 0.002);
     ConInpIT->SetName("conForward");
     if (LoadForwardWeights) ConInpIT->Load();
     else {
	  ConInpIT->ConnectFull(InitialWeights, 2,0); 
     }
//     ConInpIT->StartBinRec(InputNx/4+InputNx*InputNy/2);
     float MaxWeights=2.2;
     float LearnTauDec=20;
     float PostLearnTauDec=10;
     float BaseLine = exp(-2*StimDur/LearnTauDec);
      if (Learn) ConInpIT->SetLearnObj(
 //   	  new LearnHebbLP2(ConInpIT, 
 //   			   MaxWeights, LearnTauDec, BaseLine, LearnRate));
// 	  new LearnRossum1(ConInpIT, RossumCp, 0.003, 0.0002));
   	  new LearnHebbLP3(ConInpIT, MaxWeights, LearnTauDec, 
			   PostLearnTauDec, BaseLine, LearnRate, true));
//  	  new LearnRossum1(ConInpIT, RossumCp, 0.003, 0.00005));

     ConInpIT->Save();
     ConInpIT->SetAutoSave(false);
     MainSimLoop->AddSimElement(ConInpIT);     

    ConstSumNormalize* norm=0;
     if (Learn) {
	  norm = new ConstSumNormalize(WeightSum, true);     
	  norm->AddConnection(ConInpIT);
	  norm->NormalizeAll();
	  MainSimLoop->AddSimElement(norm);
     }

//      FiringRateNormalize* norm=0;
//      if (Learn) {
// 	  norm = new FiringRateNormalize(1.3, 0.999, 100);     
// 	  norm->AddConnection(ConInpIT);
// 	  MainSimLoop->AddSimElement(norm);
//      }

//      ConstSFNormalize* norm=0;
//      if (Learn && ConInpIT && (ITNormFrequency>0)) {
// 	  norm = new ConstSFNormalize(ITNormFrequency, 0.0005, 100);     
// 	  norm->AddConnection(ConInpIT);
// 	  MainSimLoop->AddSimElement(norm);
//      }


     connection *ConInhInpIT = 0;
     if (InitialFWInhWeights>0) {
	  float InhRossumCp=0.05*InitialFWInhWeights;
	  ConInhInpIT = new connection(BlobInputLayer, ITLayer, 1);
	  ConInhInpIT->SetName("conInhForward");
	  if (LoadForwardWeights) ConInhInpIT->Load();
	  else {
	       ConInhInpIT->ConnectFull(InitialFWInhWeights, 2,0);
	  }
	  
 	  if (InhLearnRate > 0) ConInhInpIT->SetLearnObj(
 	       new LearnHebbLP2(ConInhInpIT,
 				MaxFWInhWeights, InhLearnTauDec, InhLearnBaseLine, -InhLearnRate, false));
// 	       new LearnAntiRossum1(ConInhInpIT, InhRossumCp, 0.003, 0.001));
	  ConInhInpIT->Save();
	  ConInhInpIT->SetAutoSave(false);
	  MainSimLoop->AddSimElement(ConInhInpIT);

	  ConstSFNormalize* FWInhNorm=0;
	  if (Learn) {
	       FWInhNorm = new ConstSFNormalize(15, -0.001, 40);     
	       FWInhNorm->AddConnection(ConInhInpIT);
	       MainSimLoop->AddSimElement(FWInhNorm);
	  }


     }


     connection *ConITFBInh = 0;
//      if (FBInhWeight > 0)
//      {
// 	  ConITFBInh = new connection(ITLayer, BlobInputLayer, 1);
// 	  ConITFBInh->SetName("conFBInh");
// 	  ConITFBInh->ConnectFull(FBInhWeight, 2,0); 
// 	  float LearnInhTauDec=30;
// 	  if (Learn) ConITFBInh->SetLearnObj(
// 	       new LearnFBInh(ConITFBInh, 
// 			      FBInhMaxWeight, LearnInhTauDec,LearnInhTauDec, FBInhLearnRate, Cdep));
// //  	  new LearnHebbLP2(ConITFBInh, 
// //  			   MaxWeights, LearnInhTauDec, 0.5, LearnRate));
// //	  new LearnRossum1(ConITFBInh, RossumCp, 0.003, 0.0002));

// 	  ConITFBInh->Save();
// 	  ConITFBInh->SetAutoSave(false);
// 	  MainSimLoop->AddSimElement(ConITFBInh);     

// //      ConstSumNormalize* norm=0;
// //      if (Learn) {
// // 	  norm = new ConstSumNormalize(WeightSum, false);     
// // 	  norm->AddConnection(ConITFBInh);
// // 	  norm->NormalizeAll();
// // 	  MainSimLoop->AddSimElement(norm);
// //      }

//      }








////////// Setup Input

     float l1inpStimDur=StimDur;
     float Radius=0.2;

     bool Tune=false;
     input* l1inp;     

     enum InputType {InpHighOverlap, InpNoOverlap, OnePicture};
     InputType CurInputType=OnePicture;
//     InputType CurInputType=InpNoOverlap;

     

     switch (CurInputType) {
     case InpHighOverlap:
     {
	  PictureSequenceInput* TuneInp = new PictureSequenceInput(BlobInputLayer, 0,istrength, l1inpStimDur, 1*l1inpStimDur);
	  vector<vector<float> > StimulusArray;
	  int NStim=4;
	  for (int i=0;i<NStim;++i) {
	       vector<float> CurStimArray(InputNx*InputNy);
	       for (int xx=0;xx<InputNx;++xx) 
		    for (int yy=0;yy<InputNy/2;++yy) 
			 CurStimArray[xx + InputNx*yy]=1;
	       for (int yy=InputNy/2;yy<InputNy;++yy) 
		    for (int xx=InputNx*i/NStim;xx<InputNx*(i+1)/NStim;++xx) 
			 CurStimArray[xx + InputNx*yy]=1;
	       StimulusArray.push_back(CurStimArray);
	  }

	  TuneInp->SetPictureArray(StimulusArray);
	  TuneInp->SetName("HighOverlapInput");
//	  TuneInp->SetTestMode(l1inpStimDur, 2*l1inpStimDur);
	  TuneInp->SetMode(csimInputRandom);
//     TuneInp->SetMode(csimInputTest);
	  MainSimLoop->AddSimElement(TuneInp);
	  l1inp=TuneInp;


     }
     break;
     case InpNoOverlap:
     {
	  PictureSequenceInput* TuneInp = new PictureSequenceInput(BlobInputLayer, 0,istrength, l1inpStimDur, 1*l1inpStimDur);
	  vector<vector<float> > StimulusArray;
	  int NStim=4;
	  for (int i=0;i<NStim;++i) {
	       vector<float> CurStimArray(InputNx*InputNy);
	       for (int yy=InputNy/2;yy<InputNy;++yy) 
		    for (int xx=InputNx*i/NStim;xx<InputNx*(i+1)/NStim;++xx) 
			 CurStimArray[xx + InputNx*yy]=1;
	       StimulusArray.push_back(CurStimArray);
	  }

	  TuneInp->SetPictureArray(StimulusArray);
	  TuneInp->SetName("HighOverlapInput");
//	  TuneInp->SetTestMode(l1inpStimDur, 0.2*l1inpStimDur);
//     TuneInp->SetMode(csimInputTest);
	  TuneInp->SetMode(csimInputRandom);
	  MainSimLoop->AddSimElement(TuneInp);
	  l1inp=TuneInp;
     }	       
     break;
     case OnePicture:
     {
	  	       PictureSequenceInput* WinkelInput = new PictureSequenceInput(BlobInputLayer, 0,istrength, l1inpStimDur, IsiDur);
	       WinkelInput->SetPictureArray(&MyMovie);
	       WinkelInput->SetName("WinkelInput");
	       WinkelInput->SetMode(csimInputRandom);
//	       WinkelInput->SetRandomNext(true);
	       MainSimLoop->AddSimElement(WinkelInput);
	       l1inp = WinkelInput;

     }
     break;
     default: cout << "nothing";
     }


////////////////////////////////////

     int trial, step;     
     sec =0;
         
     if (TestSim) {
 	  probesteps=200;
 	  learnsteps= 0; // 80;
 	  MaxTrials=1;
	  cout << "Test (recall) ----------------------Simulation\n";
//	  l1inp->SetTestMode(100,1000); 
     } else cout << "NOtestsim------------------------\n";
     fflush(stdout);

     int TotalTime=0;


     MainSimLoop->Hallo();
     MainSimLoop->SaveSimInfo();
     for (trial=0; trial<MaxTrials; ++trial)
     {
	  cout << "TrialNr=" << trial << "\n";
//	  l1inp->TurnOn();
	  // learn
	  for(step=0;step<learnsteps;++step)
	  {
	       cpu_end=clock();
	       cpu_time_used = ((double) (cpu_end - cpu_start)) / CLOCKS_PER_SEC;
	       cout << "T" << trial << "  Step=" << step 
		    << " sec= " << dt*(sec++) << "  cputime=" 
		    << cpu_time_used << " sec \n  "; fflush(stdout);

	       for (t=0;t<1000;t++) 				// simulation of 1 sec
	       {
		    MainSimLoop->proceede(1000*step+t);
	       }	       
	       MainSimLoop->prepare(step);

	  }
 	  // SaveWeights
 	       ConInpIT->Save(trial);
	       if (ConITFbInh) ConITFbInh->Save(trial);
	       if (ConITSelfInh) ConITSelfInh->Save(trial);
	       if (ConInhInpIT) ConInhInpIT->Save(trial);
	  // probe
	  // change input, turn off learning, turn off normalization
// 	  cout << "probe\n";
//  	  if (Learn) {
// 	       ConInpIT->SetLearn(false);
// 	       if (ConInhInpIT) ConInhInpIT->SetLearn(false);
// 	       if (norm) norm->TurnOff();
// 	  }


// 	  MainSimLoop->SetSimTag("TEST");
// 	  for(step=0;step<probesteps;++step)
// 	  {
// 	       cpu_end=clock();
// 	       cpu_time_used = ((double) (cpu_end - cpu_start)) / CLOCKS_PER_SEC;
// 	       cout << "sec= " << dt*(sec++) << "cputime=" << cpu_time_used << " sec \n  "; fflush(stdout);
// 	       for (t=0;t<1000;t++) 				// simulation of 1 sec
// 	       {
// 		    MainSimLoop->proceede(1000*step+t);
// 	       }	       
// 	       MainSimLoop->prepare(step);
// 	  }	
//  	  if (Learn) {
// 	       ConInpIT->SetLearn(true);
// 	       if (ConInhInpIT) ConInhInpIT->SetLearn(true);
// 	       if (norm) norm->TurnOn();
// 	  }
	  MainSimLoop->SetSimTag("");
     }

     ConInpIT->Save();
     if (ConITFbInh) ConITFbInh->Save();
     if (ConITSelfInh) ConITSelfInh->Save();
     if (ConInhInpIT) ConInhInpIT->Save();
     
//      delete Layer1;
//      delete Layer2;
//      delete l1l2con;
//      delete l1inp;
     cpu_end=clock();
     cpu_time_used = ((double) (cpu_end - cpu_start)) / CLOCKS_PER_SEC;
     cout << "\n\n***********CpuTimeUsed= " << cpu_time_used << " seconds \n";
     system("bluetclbutton v4learn finished &");
}