DistributionMultiVariateNormal::DistributionMultiVariateNormal(F32 mux,F32 muy,F32 sigmax,F32 sigmay,F32 rho)
:DistributionMultiVariate(),_sigma(2,2),_mean(2),_standard_normal(0,1)
{
_mean(0)=mux;
_mean(1)=muy;
_sigma(0,0)=sigmax*sigmax; _sigma(0,1)=sigmax*sigmay*rho;
_sigma(1,0)=sigmax*sigmay*rho;_sigma(1,1)=sigmay*sigmay;
_sigma_minus_one=LinearAlgebra::inverseGaussianElimination(_sigma);
_determinant_sigma = _sigma.determinant();
_a = LinearAlgebra::AATransposeEqualMDecomposition(_sigma);
}
EXPORT
char *meanRating(void) {
return _char_rating(_mean());
}
void petabricks::PetabricksRuntime::runGraphTemplate(){
int i=0;
for(; _main!=NULL; _main=_main->nextTemplateMain(),++i){
double avg = runTrial(GRAPH_MAX_SEC, ACCTRAIN);
if(theAccuracies.empty()){
printf("%d %.6f\n", i, avg);
}else{
printf("%d %.6f %.6f %.6f\n", i, avg, _main->accuracyTarget(), _mean(theAccuracies));
theAccuracies.clear();
}
}
}
void petabricks::PetabricksRuntime::runGraphMode(){
for(int n=GRAPH_MIN; n<=GRAPH_MAX; n=_incN(n)){
setSize(n);
double avg = runTrial(GRAPH_MAX_SEC, ACCTRAIN);
if(avg<std::numeric_limits<double>::max()){
if(theAccuracies.empty()){
printf("%d %.6f\n", n, avg);
}else{
printf("%d %.6f %.6f\n", n, avg, _mean(theAccuracies));
theAccuracies.clear();
}
}
fflush(stdout);
if(avg > GRAPH_MAX_SEC) break;
}
}
void petabricks::PetabricksRuntime::runGraphParamMode(const std::string& param){
jalib::JTunable* tunable = jalib::JTunableManager::instance().getReverseMap()[param];
JASSERT(tunable!=0)(param).Text("parameter not found");
GRAPH_MIN = std::max(GRAPH_MIN, tunable->min().i());
GRAPH_MAX = std::min(GRAPH_MAX, tunable->max().i());
for(int n=GRAPH_MIN; n<=GRAPH_MAX; n=_incN(n)){
tunable->setValue(n);
double avg = runTrial(GRAPH_MAX_SEC, ACCTRAIN);
if(avg<std::numeric_limits<double>::max()){
if(theAccuracies.empty()){
printf("%d %.6f\n", n, avg);
}else{
printf("%d %.6f %.6f\n", n, avg, _mean(theAccuracies));
theAccuracies.clear();
}
}
fflush(stdout);
if(avg > GRAPH_MAX_SEC) break;
}
}
