void NeuralPatternArray::read(TextFileReader const& _file) throw()
{
TextFileReader file = _file;
Text text = file.readLine();
if(text.contains("NeuralPatternArray:v1"))
{
//printf("is valid NeuralPatternArray file\n");
while(file.isEof() == false)
{
NeuralPattern pat;
pat.read(file);
if((pat.getNumInputs() > 0) && (pat.getNumOutputs() > 0))
{
//printf("..adding a pattern..\n");
this->add(pat);
}
}
}
}
void NeuralPattern::read(TextFileReader const& _file) throw()
{
TextFileReader file = _file;
Text text;
int line = 0;
int numInputs = -1, numOutputs = -1;
NumericalArray<float> inputVector;
NumericalArray<float> outputVector;
while((text = file.readLine()).size() != 0)
{
if(line == 0)
{
// should be the format/version string
if(text.contains("NeuralPatternSimple:v1") == false)
{
printf("NeuralPattern::read wrong version type\n");
return;
}
}
else if(numInputs < 0)
{
// should be the number of inputs
sscanf(text.getArray(), "Inputs:%d", &numInputs);
if(numInputs < 1)
{
printf("NeuralPattern::read invalid number of inputs\n");
return;
}
//printf("ins: %d\n", numInputs);
inputVector = NumericalArray<float>::withSize(numInputs);
}
else if(numInputs > 0)
{
int index;
float value;
sscanf(text.getArray(), "%d %f", &index, &value);
numInputs--;
//printf("in[%3d] = %f\n", index, value);
inputVector[index] = value;
}
else if(numOutputs < 0)
{
// should be the number of outputs
sscanf(text.getArray(), "Outputs:%d", &numOutputs);
if(numOutputs < 1)
{
printf("NeuralPattern::read invalid number of numOutputs\n");
return;
}
//printf("outs: %d\n", numOutputs);
outputVector = NumericalArray<float>::withSize(numOutputs);
}
else if(numOutputs > 0)
{
int index;
float value;
sscanf(text.getArray(), "%d %f", &index, &value);
numOutputs--;
//printf("out[%3d] = %f\n", index, value);
outputVector[index] = value;
}
if((numInputs == 0) && (numOutputs == 0))
{
//printf("CREATING PATTERN\n");
setInternal(new NeuralPatternSimpleInternal(inputVector, outputVector));
return;
}
line++;
}
}
void NeuralNetwork::read(TextFileReader const& _file) throw()
{
TextFileReader file = _file;
Text text;
int line = 0;
int numLayers = -1;
float learnRate, actFuncOffset;
IntArray nodes;
while(file.isEof() == false)
{
text = file.readLine();
if(line == 0)
{
// should be the format/version string
if(text.contains("NeuralNetworkSimple:v1") == false)
{
printf("NeuralNetwork::read wrong version type\n");
return;
}
}
else if(line == 1)
{
// should be learn rate and act func offset
sscanf(text.getArray(), "learnRate: %f actFuncOffset: %f", &learnRate, &actFuncOffset);
}
else if(line == 2)
{
// should be the number of layers
sscanf(text.getArray(), "Layers:%d", &numLayers);
if(numLayers < 2)
{
printf("NeuralNetwork::read invalid number of layers\n");
return;
}
nodes = IntArray::withSize(numLayers, false);
}
else if(numLayers > 0)
{
int layer, numNodes;
sscanf(text.getArray(), "Layer %d:%d", &layer, &numNodes);
numLayers--;
nodes[layer] = numNodes;
if(numLayers == 0)
{
if(nodes == getStructure())
{
setLearnRate(learnRate);
setActFuncOffset(actFuncOffset);
}
else
{
setInternal(new NeuralNetworkSimpleInternal(nodes, learnRate, actFuncOffset));
}
}
}
else
{
// it's a weight or a threshold
int layer, node, index;
float value;
sscanf(text.getArray(), "%d %d %d %f", &layer, &node, &index, &value);
if(index < 0)
setThreshold(layer, node, value);
else
setWeight(layer, node, index, value);
}
line++;
}
}