void read( const FileNode& fn )
{
clear();
vector<int> _layer_sizes;
readVectorOrMat(fn["layer_sizes"], _layer_sizes);
setLayerSizes( _layer_sizes );
int i, l_count = layer_count();
read_params(fn);
size_t esz = weights[0].elemSize();
FileNode w = fn["input_scale"];
w.readRaw("d", weights[0].ptr(), weights[0].total()*esz);
w = fn["output_scale"];
w.readRaw("d", weights[l_count].ptr(), weights[l_count].total()*esz);
w = fn["inv_output_scale"];
w.readRaw("d", weights[l_count+1].ptr(), weights[l_count+1].total()*esz);
FileNodeIterator w_it = fn["weights"].begin();
for( i = 1; i < l_count; i++, ++w_it )
(*w_it).readRaw("d", weights[i].ptr(), weights[i].total()*esz);
trained = true;
}
ANN_MLPImpl()
{
clear();
setActivationFunction( SIGMOID_SYM, 0, 0 );
setLayerSizes(Mat());
setTrainMethod(ANN_MLP::RPROP, 0.1, FLT_EPSILON);
}
void MultilayerPerceptron::initMLP(int ninputs, int noutputs, const vector<int> &hiddenLayerSizes, const TransferFunctionType &tf)
{
tres = new MLPTrainingResult();
nInputs = ninputs;
setAlfa(1);
setLayerSizes(hiddenLayerSizes);
setInputSize(ninputs);
setOutputSize(noutputs);
setTransferFunctionType(tf);
// setOutputType(Continuous);
randomizeWeights();
connect(this, SIGNAL(finished()), SLOT(finished()));
}
