// Ntk Construction Functions
void
V3BvNtk::initialize() {
assert (!_inputData.size());
// Create Constant BV_CONST = 1'b0 for Sync with AIG_FALSE
const V3NetId id = createNet(1); assert (!id.id);
_inputData.back().push_back(0); createConst(id);
}
NeuralNet::NeuralNet() {
numInputs = Params::numInputs;
numOutputs = Params::numOutputs;
numHiddenLayers = Params::numHidden;
neuronsPerHiddenLayer = Params::neuronsPerHiddenLayer;
createNet();
}
BlisSolution *
VrpModel::userFeasibleSolution(const double *solution, bool &userFeasible)
{
double objValue = 0.0;
CoinPackedVector *solVec = getSolution(solution);
VrpSolution *vrpSol = NULL;
int msgLevel = AlpsPar_->entry(AlpsParams::msgLevel);
userFeasible = true;
createNet(solVec);
if (!n_->isIntegral_){
if (msgLevel > 200) {
std::cout << "UserFeasible: not integral" << std::endl;
}
userFeasible = false;
}
else {
int rcnt = n_->connected();
int i;
for (i = 0; i < rcnt; i++){
if (n_->compCuts_[i+1] < 2 - etol_){
userFeasible = false;
if (msgLevel > 200) {
std::cout << "UserFeasible: not 2" << std::endl;
}
break;
}
else if (n_->compDemands_[i+1] > capacity_){
userFeasible = false;
if (msgLevel > 200) {
std::cout << "UserFeasible: greater than capacity" << std::endl;
}
break;
}
}
}
if (userFeasible) {
// Compute obj value
for (int k = 0; k < numCols_; ++k) {
objValue += objCoef_[k] * solution[k];
}
// Create a VRP solution
vrpSol = new VrpSolution(getNumCols(),
solution,
objValue,
this);
// TODO: add tour
}
// Free memory.
delete solVec;
return vrpSol;
}
RaceClient::RaceClient(QString host, int port, QObject *parent)
{
m_host = host;
m_port = port;
m_socket = new QTcpSocket(this);
m_state = RaceClient::Connecting;
createNet(false);
emit stateChanged(m_state);
connect(m_socket, SIGNAL(readyRead()),
this, SLOT(readFromServer()));
connect(m_socket, SIGNAL(connected()),
this, SLOT(onConnected()));
connect(m_socket, SIGNAL(error(QAbstractSocket::SocketError)),
this, SLOT(onError(QAbstractSocket::SocketError)));
}
void Signatures::createSets(vector<Node*> vNodes){
createNet(vNodes);
createNetPred(vNodes);
createA_Sig(vNodes);
}
NNet::NNet( int numberOfInputs, int numberOfOutputs, std::vector<int> layerSize )
{
setMemory( false );
createNet( numberOfInputs, numberOfOutputs, layerSize );
}
int main() {
double inputs[MAX_NO_OF_INPUTS];
double outputTargets[MAX_NO_OF_OUTPUTS];
/* determine layer paramaters */
int noOfLayers = 2; // input layer excluded
int noOfNeurons[] = {10,1};
int noOfInputs[] = {2,10};
char axonFamilies[] = {'g','l'};
double actFuncFlatnesses[] = {1,1,1};
createNet(noOfLayers, noOfNeurons, noOfInputs, axonFamilies, actFuncFlatnesses, 1);
/* train it using batch method */
int i;
double tempTotal1, tempTotal2;
int counter = 0;
for(i = 0; i < TRAINING_ITERATION; i++) {
inputs[0] = getRand();
inputs[1] = getRand();
inputs[2] = getRand();
inputs[3] = getRand();
tempTotal1 = inputs[0] + inputs[1];
tempTotal2 = inputs[2] - inputs[3];
// tempTotal = inputs[0] + inputs[1];
feedNetInputs(inputs);
updateNetOutput();
// outputTargets[0] = sin(tempTotal1)*2+0.5*exp(tempTotal2)-cos(inputs[1]+inputs[3])/2;
outputTargets[0] = sin(tempTotal1)*2+0.5*exp(tempTotal2);
// outputTargets[1] = (double)cos(tempTotal);
/* train using batch training ( don't update weights, just cumulate them ) */
//trainNet(0, 0, 1, outputTargets);
trainNet(0, 0, 1, outputTargets);
counter++;
/* apply batch changes after 100 loops use .8 learning rate and .8 momentum */
if(counter == 100) { applyBatchCumulations(.3,.3); counter = 0;} //!~~~swd: should be within (0,1)
}
/* test it */
double *outputs;
double target_out[50];
double actual_out[50];
printf("Sin Target \t Output \n");
printf("---------- \t -------- \t ---------- \t --------\n");
for(i = 0; i < 50; i++) {
inputs[0] = getRand();
inputs[1] = getRand();
inputs[2] = getRand();
inputs[3] = getRand();
tempTotal1 = inputs[0] + inputs[1];
tempTotal2 = inputs[2] - inputs[3];
target_out[i] = sin(tempTotal1)*2+0.5*exp(tempTotal2);
// target_out[i] = sin(tempTotal1)*2+0.5*exp(tempTotal2)-cos(inputs[1]+inputs[3])/2;
feedNetInputs(inputs);
updateNetOutput();
outputs = getOutputs();
actual_out[i] = outputs[0];
printf( "%f \t %f \n", target_out[i], actual_out[i]);
}
float Rsquared_ans=Rsquared(target_out, actual_out, 50);
printf("Result Summary: \n");
printf("The Rsquared Value is : %f \n", Rsquared_ans);
printf("finish!!!\n");
//getch();
return 0;
}
// Ntk Construction Functions
void
V3Ntk::initialize() {
assert (!_inputData.size());
// Create Constant AIG_FALSE
const V3NetId id = createNet(1); assert (!id.id); createConst(id);
}
