static status
deleteCellTable(Table tab, TableCell cell, BoolObj keep)
{ if ( cell->layout_manager == (LayoutManager)tab )
{ int tx = valInt(cell->column) + valInt(cell->col_span);
int ty = valInt(cell->row) + valInt(cell->row_span);
int x, y;
removeCellImageTable(tab, cell, keep);
for(y=valInt(cell->row); y<ty; y++)
{ TableRow row = getRowTable(tab, toInt(y), OFF);
if ( row )
{ for(x=valInt(cell->column); x<tx; x++)
elementVector((Vector)row, toInt(x), NIL);
}
}
assign(cell, layout_manager, NIL);
changedTable(tab);
requestComputeLayoutManager((LayoutManager)tab, DEFAULT);
}
succeed;
}
static status
appendTable(Table tab, TableCell cell, Int x, Int y)
{ int cspan = valInt(cell->col_span);
int rspan = valInt(cell->row_span);
int dy;
if ( isDefault(x) )
x = tab->current->x;
if ( isDefault(y) )
y = tab->current->y;
if ( notNil(tab->device) && notNil(cell->image) )
send(tab->device, NAME_display, cell->image, EAV);
assign(cell, layout_manager, tab);
assign(cell, column, x);
assign(cell, row, y);
for(dy=0; dy<rspan; dy++)
{ Int cy = toInt(valInt(y)+dy);
TableRow row = getRowTable(tab, cy, ON);
int dx;
for(dx=0; dx<cspan; dx++)
{ Int cx = toInt(valInt(x)+dx);
cellTableRow(row, cx, cell);
}
}
advance_table(tab);
requestComputeLayoutManager((LayoutManager)tab, DEFAULT);
return changedTable(tab);
}
//****************************************************************************
//****************************************************************************
//****************************************************************************
// RunKPLS - run kpls to classify objects
//****************************************************************************
void PatternAnalysisWizard::runKPLS()
{
#ifdef USE_KPLS
//Find out which features are checked (which columns to use).
std::vector<int> columnsToUse;
QList<QAbstractButton *> buttons = featureGroup->buttons();
for(int b = 0; b<buttons.size(); ++b)
{
if( buttons.at(b)->isChecked() )
{
if(featureGroup->id(buttons.at(b)) != 0)
{
columnsToUse.push_back( featureGroup->id(buttons.at(b)) );
}
else
{
for(int col=0; col<(int)m_table->GetNumberOfColumns(); ++col)
{
std::string col_name = m_table->GetColumnName(col);
if(col_name.find("Zern")!=std::string::npos)
{
columnsToUse.push_back(col);
}
}
}
}
}
if(columnsToUse.size() <= 0)
return;
this->KPLSrun(columnsToUse);
emit changedTable();
emit enableModels();
#else
QMessageBox::information(this, tr("MESSAGE"), tr("FARSIGHT was not compiled with KPLS library"));
#endif
}
static status
rowSpanTableCell(TableCell cell, Int span)
{ if ( cell->row_span != span )
{ Table tab = table_of_cell(cell);
if ( tab )
{ int x,y;
int fy = valInt(cell->row);
int h = valInt(span);
int oh = valInt(cell->row_span);
int ty = max(h,oh)+fy;
for(y=fy+1; y<ty; y++)
{ TableRow row = getRowTable(tab, toInt(y), ON);
for(x=valInt(cell->column); x<valInt(cell->column)+valInt(cell->col_span); x++)
{ Any e;
if ( y-fy < h )
e = cell;
else
e = NIL;
cellTableRow(row, toInt(x), e);
}
}
assign(cell, row_span, span);
changedTable(tab); /* changed line-pattern */
requestComputeLayoutManager((LayoutManager)tab, DEFAULT);
} else
{ assign(cell, row_span, span);
}
}
succeed;
}
static status
colSpanTableCell(TableCell cell, Int span)
{ if ( cell->col_span != span )
{ Table tab = table_of_cell(cell);
if ( tab )
{ int x,y;
int fx = valInt(cell->column);
int w = valInt(span);
int ow = valInt(cell->col_span);
int tx = max(w,ow)+fx;
for(y=valInt(cell->row); y<valInt(cell->row)+valInt(cell->row_span); y++)
{ TableRow row = getRowTable(tab, toInt(y), ON);
for(x=fx+1; x<tx; x++)
{ Any e;
if ( x-fx < w )
e = cell;
else
e = NIL;
cellTableRow(row, toInt(x), e);
}
}
assign(cell, col_span, span);
changedTable(tab); /* changed line-pattern */
requestComputeLayoutManager((LayoutManager)tab, DEFAULT);
} else
{ assign(cell, col_span, span);
}
}
succeed;
}
//****************************************************************************
//****************************************************************************
//****************************************************************************
// RunSVM - run support vector machine in one class mode to find outliers.
//****************************************************************************
void PatternAnalysisWizard::runSVM()
{
bool normalize = true; //ALWAYS TRUE!!
double nu = 0.10;
//Find out which featueres are checked (which columns to use).
std::vector<int> columnsToUse;
QList<QAbstractButton *> buttons = featureGroup->buttons();
for(int b = 0; b<buttons.size(); ++b)
{
if( buttons.at(b)->isChecked() )
{
columnsToUse.push_back( featureGroup->id(buttons.at(b)) );
}
}
//Setup the scaling values
std::vector<double> f_max; //The maximum value of each feature
f_max.assign(columnsToUse.size(), -DBL_MAX);
std::vector<double> f_min; //The minimum value of each feature
f_min.assign(columnsToUse.size(), DBL_MAX);
std::vector< std::vector< double > > features; //Will contain the normalized features
features.resize( m_table->GetNumberOfRows() );
//exract data from the model and get min/max values:
for(int r=0; r<(int)features.size(); ++r)
{
for(int c=0; c<(int)columnsToUse.size(); ++c)
{
int col = columnsToUse.at(c);
double val = m_table->GetValue(r,col).ToDouble();
features.at(r).push_back( val );
if( normalize )
{
//Also gather min/max
f_max.at(c) = val > f_max.at(c) ? val : f_max.at(c);
f_min.at(c) = val < f_min.at(c) ? val : f_min.at(c);
}
}
}
if( normalize )
{
//Normalize/Scale the Features Data:
double upper = 1;
double lower = -1;
double desired_range = upper-lower;
for(int r=0; r<(int)features.size(); ++r)
{
for(int c=0; c<(int)columnsToUse.size(); ++c)
{
double oldval = features.at(r).at(c);
features.at(r).at(c) = lower + desired_range * (oldval - f_min.at(c)) / (f_max.at(c) - f_min.at(c));
}
}
}
//Create the libSVM problem:
#define Malloc(type,n) (type *)malloc((n)*sizeof(type))
struct svm_problem prob;
prob.l = (int)features.size(); //Number of objects
prob.y = Malloc(double,prob.l); //Array Containing target values (unknowns)
prob.x = Malloc(struct svm_node *,prob.l); //Array of Pointers to Nodes
for(int r=0; r<prob.l; ++r)
{
prob.y[r] = 1; //This is the label (target) and it is unknown
struct svm_node *x_space = Malloc(struct svm_node, columnsToUse.size()+1); //Individual node
for(int c=0; c<(int)columnsToUse.size(); ++c)
{
x_space[c].index = c+1;
x_space[c].value = features.at(r).at(c);
}
x_space[ columnsToUse.size() ].index = -1;
prob.x[r] = &x_space[0]; //Point to this new set of nodes.
}
//Set the Parameters
struct svm_parameter param;
param.svm_type = ONE_CLASS;
param.kernel_type = RBF;
param.degree = 3;
param.gamma = 1.0/double(prob.l); // 1/k
//param.gamma = 1;
param.coef0 = 0;
//param.nu = 0.1;
param.nu = nu;
param.cache_size = 100;
param.C = 1;
param.eps = .001;
param.p = 0.1;
param.shrinking = 1;
param.probability = 0;
param.nr_weight = 0;
param.weight_label = NULL;
param.weight = NULL;
//Now train
struct svm_model *m_svm_model;
m_svm_model = svm_train(&prob,¶m);
svm_destroy_param(¶m);
free(prob.y);
free(prob.x);
//Predict:
std::vector<int> outliers;
for(int r=0; r<prob.l; r++)
{
struct svm_node *x = Malloc(struct svm_node,columnsToUse.size()+1); //Individual node
for(int c=0; c<(int)columnsToUse.size(); ++c)
{
x[c].index = c+1;
x[c].value = features.at(r).at(c);
}
x[ columnsToUse.size() ].index = -1;
double v = svm_predict(m_svm_model,x);
free(x);
if( v == -1 )
{
outliers.push_back( r );
}
}
svm_destroy_model(m_svm_model);
//If need to create a new column do so now:
vtkAbstractArray * output = m_table->GetColumnByName(columnForPrediction);
if(output == 0)
{
vtkSmartPointer<vtkDoubleArray> column = vtkSmartPointer<vtkDoubleArray>::New();
column->SetName( columnForPrediction );
column->SetNumberOfValues( m_table->GetNumberOfRows() );
m_table->AddColumn(column);
}
for(int row = 0; (int)row < m_table->GetNumberOfRows(); ++row) //Set all values to 0
{
m_table->SetValueByName(row,columnForPrediction, 0);
}
for(int i = 0; i < (int)outliers.size(); ++i) //Set outliers to 1
{
int row = outliers.at(i);
m_table->SetValueByName(row,columnForPrediction, 1);
}
emit changedTable();
/*
model->setHeaderData( columnForPrediction, Qt::Horizontal, tr("outlier?") );
//stop signalling:
model->blockSignals(true);
int z = 0;
int o = 1;
for(int row = 0; (int)row < model->rowCount(); ++row) //Set all values to 0
{
model->setData(model->index(row, columnForPrediction), z);
}
for(int i = 0; i < (int)outliers.size()-1; ++i) //Set outliers to 1
{
model->setData(model->index(outliers.at(i), columnForPrediction), o);
}
//turn signals back on & change one more piece of data to force dataChanged signal
model->blockSignals(false);
model->setData(model->index(outliers.back(), columnForPrediction), o);
*/
}
