void CSvmModel::RunSvmPredict( const string& strDataFile, const string& strModelFile, REAL& rOutput )
{
//the svm prediction command is,
//Usage: svm-predict [options] test_file model_file output_file
char buf[_MAX_LINE];
ostrstream ostr( buf, ELEMENTS(buf) );
ostr<<SVM_PREDICT_CMD<<" "<<strDataFile.c_str()<<" "<<strModelFile.c_str()<<" "<<SVM_PREDICT_FILE<<endl;
system( buf );
ifstream ifg( SVM_PREDICT_FILE );
ifg>>rOutput;
}
void CHostSet::LoadFromFile( const char* strFile )
{
ifstream ifg(strFile);
char buf[MAXLINE];
while( !ifg.eof() && !ifg.fail() ){
ifg>>buf;
if( strncmp(buf, "//", 2)==0 )continue;
ParseHostName( buf );
// hostent* phent = gethostbyname( buf );
// if( phent!=NULL ){
// IN_ADDR addr;
// bcopy( phent->h_addr, &addr, sizeof(IN_ADDR) );
// Add( addr, true );
// }
}
}
void CSvmModel::TransformToSvmData( const string& strDataFile, int nInputs, int nOutputs, int nPredId, const string& strOutput )
{
vector<REAL> vcInputs(nInputs);
vector<REAL> vcOutputs(nOutputs);
ifstream ifg( strDataFile.c_str() );
ofstream ofg( strOutput.c_str() );
while( !ifg.eof() && !ifg.fail() ){
//read all the input and input vectors at each line
for( int i=0; i<nInputs; i++ )ifg>>vcInputs[i];
for( i=0; i<nOutputs; i++ )ifg>>vcOutputs[i];
if( ifg.eof() || ifg.fail() )break;
TransformSvmLine( ofg, vcInputs, vcOutputs, nPredId );
}
}
void CRegionalMetaModel::ReadTrainingData( const string& strDataFile, int nInputs, int nOutputs, vector< vector<REAL> >& vcInputs, vector< vector<REAL> >& vcOutputs )
{
ifstream ifg( strDataFile.c_str() );
char buf[_MAX_LINE];
while( true ){
ifg.getline( buf, ELEMENTS(buf) );
if( ifg.fail() || ifg.eof() )break;
istrstream istr(buf);
vector<REAL> vci(nInputs);
for( int i=0; i<nInputs; i++ )istr>>vci[i];
vector<REAL> vco(nOutputs);
for( i=0; i<nOutputs; i++ )istr>>vco[i];
vcInputs.push_back( vector<REAL>() );
vcInputs.back().swap( vci );
vcOutputs.push_back( vector<REAL>() );
vcOutputs.back().swap( vco );
}
}
void CNeuralModel::Train( const string& strTrainData, const string& strValidData )
{
if( m_pNeuralNet!=NULL ){
delete m_pNeuralNet;
m_pNeuralNet = NULL;
}
double rValidRatio = 0.3;
int count = 0;
char buf[_MAX_LINE];
{
std::ifstream ifg( strTrainData.c_str() );
std::ofstream ofgTrain( TRAIN_FILE );
std::ofstream ofgValid( VALID_FILE );
ifg.getline( buf, ELEMENTS(buf) );
while( !ifg.eof() && !ifg.fail() ){
trimleft( buf );
if( buf[0]!=NULL ){
count++;
if( count%4 == 0 ){
// if( uni_rand()<rValidRatio ){
ofgValid<<buf<<endl;
}else{
ofgTrain<<buf<<endl;
}
}
ifg.getline( buf, ELEMENTS(buf) );
}
}
m_pNeuralNet = new CMiniNeuralNet();
double rError = RunMatlabNet( NET_CMD, &m_vcLayerNodes[0], m_vcLayerNodes.size(), &m_vcTrainBias[0],
TRAIN_FILE, MODEL_FILE, VALID_FILE );
cout<<"Error: "<<rError<<endl;
m_pNeuralNet->Create( &m_vcLayerNodes[0], m_vcLayerNodes.size(), NeuroTanSig );
m_pNeuralNet->LoadMatlabNet( MODEL_FILE );
}
void fault_tree_graph::print_FTAs()
{
std::set<ft_vertex_t> tasks;
std::pair< ft_vertex_iter,ft_vertex_iter> ft_vtxs;
for (ft_vtxs = boost::vertices(ig); ft_vtxs.first != ft_vtxs.second; ++ft_vtxs.first)
if (ig[*ft_vtxs.first].layer == Layer::TASK)
tasks.insert(*ft_vtxs.first);
//loop:
for (std::set<ft_vertex_t>::iterator it = tasks.begin();it != tasks.end();++it)
{
typedef boost::property_map<FT_Graph, boost::vertex_color_t>::type color_map_t;
color_map_t colorMap = get(boost::vertex_color, ig);; //Create a color map
PRINT_DEBUG("pre bfs step with node: "+ ig[*it].name);
#ifdef DEBUG
FT_Graph::vertex_iterator ita, itEnd;
for (boost::tie(ita, itEnd) = boost::vertices(ig); ita != itEnd; ita++)
{
std::cout << "Color of node prebfs " << *ita << " is " << colorMap[*ita] << std::endl;
}
#endif
FTA_visitor vis = FTA_visitor(*it);
try
{
boost::breadth_first_search(ig, *it,boost::color_map(colorMap).visitor(vis) );
}
catch (int exception)
{
if (exception == 1)
PRINT_DEBUG("found");
}
#ifdef DEBUG
for (boost::tie(ita, itEnd) = boost::vertices(ig); ita != itEnd; ita++)
{
std::cout << "Color of node " << *ita << " is " << colorMap[*ita] << std::endl;
}
#endif
//visit
PRINT_DEBUG("bfs step passed");
boost::filtered_graph<FT_Graph,true_edge_predicate<FT_Graph>,FT_print_filter_c> ifg (ig,true_edge_predicate<FT_Graph>(ig),FT_print_filter_c(ig,colorMap));
PRINT_DEBUG("filter step passed");
//filter
std::ofstream myfile;
myfile.open ("/home/emanuele/Documents/tmp_graph/"+ig[*it].name+".dot");
boost::write_graphviz(myfile, ig,make_vertex_writer(boost::get(&FT_Node::crit, ig),boost::get (&FT_Node::name, ig))); //should use a different template: crit is mapped on the first three colors of layers
myfile.close();
PRINT_DEBUG("print step passed");
//print
FT_Graph::vertex_iterator it2, it2End;
for (boost::tie(it2, it2End) = boost::vertices(ig); it2 != it2End; it2++)
{
colorMap[*it2] = boost::default_color_type::white_color;
}
}
}
