void ViTrainer::clear()
{
setNetwork(NULL);
setLearningRate(DEFAULT_LEARNING_RATE);
setIterationLimit(0); //Unlimited
setErrorLimit(0);
mCurrentIteration = -1;
mCurrentError = 1;
viDeleteAll(mErrorFunctions);
}
bool ViTrainer::importData(ViElement element)
{
if(element.name() != "trainer")
{
element = element.child("trainer");
if(element.name() != "trainer")
{
return false;
}
}
ViElement theName = element.child("name");
if(theName.isNull())
{
return false;
}
if(theName.toString() != name("Trainer"))
{
return false;
}
ViElement limit = element.child("iterationlimit");
if(limit.isNull())
{
LOG("The iteration limit could not be imported", QtCriticalMsg);
}
else
{
setIterationLimit(limit.toInt());
}
limit = element.child("errorlimit");
if(limit.isNull())
{
LOG("The error limit could not be imported", QtCriticalMsg);
}
else
{
setErrorLimit(limit.toReal());
}
limit = element.child("learningrate");
if(limit.isNull())
{
LOG("The learning could not be imported", QtCriticalMsg);
}
else
{
setLearningRate(limit.toReal());
}
addErrorFunction(ViErrorFunctionManager::create(element));
return true;
}
void ListModel::applyParams(SolverParams ¶ms) {
LOG_I("Applying custom parameters");
if (ls.getNbPhases() == 0) {
auto phase = ls.createPhase();
if(params.iterLimit != -1)
phase.setIterationLimit(static_cast<long long>(params.iterLimit));
if(params.timeLimit != -1.0)
phase.setTimeLimit(static_cast<int>(params.timeLimit));
auto param = ls.getParam();
param.setNbThreads(params.threadCount);
param.setSeed(params.seed);
param.setVerbosity(params.verbosityLevel);
if (params.traceobj) {
int timeBetweenDisplays = static_cast<int>(ceil(MSECS_BETWEEN_TRACES_LONG / 1000.0));
param.setTimeBetweenDisplays(timeBetweenDisplays);
}
}
}
void ViTrainer::trainSingle()
{
setIterationLimit(1);
train();
}
void ViTrainer::train(ViNeuralNetwork *network, int iterationLimit, qreal errorLimit)
{
setIterationLimit(iterationLimit);
setErrorLimit(errorLimit);
train(network);
}
