Constraint::Constraint()
: Value(0.0),
Type(None),
AlignmentType(Undef),
Name(""),
First(GeoUndef),
FirstPos(none),
Second(GeoUndef),
SecondPos(none),
Third(GeoUndef),
ThirdPos(none),
LabelDistance(10.f),
LabelPosition(0.f),
isDriving(true),
InternalAlignmentIndex(-1),
isInVirtualSpace(false)
{
// Initialize a random number generator, to avoid Valgrind false positives.
static boost::mt19937 ran;
static bool seeded = false;
if (!seeded) {
ran.seed(QDateTime::currentMSecsSinceEpoch() & 0xffffffff);
seeded = true;
}
static boost::uuids::basic_random_generator<boost::mt19937> gen(&ran);
tag = gen();
}
void UPGMpp::getRandomAssignation(CGraph &graph,
map<size_t,size_t> &assignation,
TInferenceOptions &options )
{
static boost::mt19937 rng1;
static bool firstExecution = true;
if ( firstExecution )
{
rng1.seed(std::time(0));
firstExecution = false;
}
vector<CNodePtr> &nodes = graph.getNodes();
for ( size_t node = 0; node < nodes.size(); node++ )
{
// TODO: Check if a node has a fixed value and consider it
size_t N_classes = nodes[node]->getType()->getNumberOfClasses();
boost::uniform_int<> generator(0,N_classes-1);
int state = generator(rng1);
assignation[nodes[node]->getID()] = state;
}
/*map<size_t,size_t>::iterator it;
for ( it = assignation.begin(); it != assignation.end(); it++ )
cout << "[" << it->first << "] " << it->second << endl;*/
}
std::vector<T1> generateRandomSet(const unsigned int size_,
const T1 min_,
const T1 max_,
const bool allowRepetition_)
{
generator.seed(std::rand());
#if BOOST_MINOR_VERSION <= 46
NumberType range(min_, max_);
boost::variate_generator<Generator&, NumberType> dist(generator, range);
#else
Generator dist(min_, max_);
#endif
std::vector<T1> numbers;
numbers.reserve(size_);
std::map<T1, bool> used;
while (numbers.size() < size_)
{
#if BOOST_MINOR_VERSION <= 46
T1 number = dist();
#else
T1 number = dist(generator);
#endif
if (allowRepetition_ || used.find(number) == used.end())
{
used[number] = true;
numbers.push_back(number);
}
}
return numbers;
}
void readSeed(boost::program_options::variables_map& variableMap, boost::mt19937& randomSource)
{
if(variableMap.count("seed") > 0)
{
randomSource.seed(variableMap["seed"].as<int>());
}
}
void setUp() {
randomGen.seed(time(NULL));
eventLoop = new DummyEventLoop();
timerFactory = new DummyTimerFactory();
connection = boost::make_shared<MockeryConnection>(failingPorts, true, eventLoop);
//connection->onDataSent.connect(boost::bind(&SOCKS5BytestreamServerSessionTest::handleDataWritten, this, _1));
//stream1 = boost::make_shared<ByteArrayReadBytestream>(createByteArray("abcdefg")));
// connection->onDataRead.connect(boost::bind(&SOCKS5BytestreamClientSessionTest::handleDataRead, this, _1));
}
void setUp() {
crypto = boost::shared_ptr<CryptoProvider>(PlatformCryptoProvider::create());
destination = "092a44d859d19c9eed676b551ee80025903351c2";
randomGen.seed(static_cast<unsigned int>(time(NULL)));
eventLoop = new DummyEventLoop();
timerFactory = new DummyTimerFactory();
connection = boost::make_shared<MockeryConnection>(failingPorts, true, eventLoop);
//connection->onDataSent.connect(boost::bind(&SOCKS5BytestreamServerSessionTest::handleDataWritten, this, _1));
//stream1 = boost::make_shared<ByteArrayReadBytestream>(createByteArray("abcdefg")));
// connection->onDataRead.connect(boost::bind(&SOCKS5BytestreamClientSessionTest::handleDataRead, this, _1));
}
void init ()
{
std::chrono::high_resolution_clock::time_point now =
std::chrono::high_resolution_clock::now();
std::chrono::nanoseconds time =
std::chrono::duration_cast<std::chrono::nanoseconds>
(now.time_since_epoch () );
ran.seed (time.count() );
pid = getpid();
}
int Utils::getRandomNumber(const int min_,
const int max_)
{
generator.seed(std::time(0));
#if BOOST_MINOR_VERSION <= 46
boost::uniform_int<> range(min_, max_);
boost::variate_generator<boost::mt19937&, boost::uniform_int<> > dist(generator, range);
return dist();
#else
boost::random::uniform_int_distribution<> dist(min_, max_);
return dist(generator);
#endif
}
void parse_args(int argc, char* argv[]) {
uint32_t seed = 0;
bool has_seed = false;
struct option long_opts[] = {
{ "help", false, nullptr, 'h' },
{ "seed", true, nullptr, 's' },
{ nullptr, 0, nullptr, 0 }
};
while (true) {
optopt = 1;
int optchar = getopt_long(argc, argv, "hs:", long_opts, nullptr);
if (optchar == -1) {
break;
}
switch (optchar) {
case 's': {
char *endptr;
seed = strtol(optarg, &endptr, 0);
if (endptr == optarg || *endptr != '\0') {
fprintf(stderr, "invalid seed value \"%s\": must be a positive "
"integer\n", optarg);
exit(1);
}
has_seed = true;
break;
}
case 'h':
print_usage(stdout, argv[0]);
exit(0);
case '?':
exit(1);
default:
// Only happens if someone adds another option to the optarg string,
// but doesn't update the switch statement to handle it.
fprintf(stderr, "unknown option \"-%c\"\n", optchar);
exit(1);
}
}
if (!has_seed) {
seed = time(nullptr);
}
printf("seed: %" PRIu32 "\n", seed);
rng.seed(seed);
}
boost::unit_test::test_suite* init_unit_test_suite(int argc, char* argv[]) {
THRIFT_UNUSED_VARIABLE(argc);
THRIFT_UNUSED_VARIABLE(argv);
uint32_t seed = static_cast<uint32_t>(time(NULL));
printf("seed: %" PRIu32 "\n", seed);
rng.seed(seed);
boost::unit_test::test_suite* suite = &boost::unit_test::framework::master_test_suite();
suite->p_name.value = "ZlibTest";
uint32_t buf_len = 1024 * 32;
add_tests(suite, gen_uniform_buffer(buf_len, 'a'), buf_len, "uniform");
add_tests(suite, gen_compressible_buffer(buf_len), buf_len, "compressible");
add_tests(suite, gen_random_buffer(buf_len), buf_len, "random");
suite->add(BOOST_TEST_CASE(test_no_write));
return NULL;
}
bool init_unit_test_suite() {
uint32_t seed = static_cast<uint32_t>(time(NULL));
#ifdef HAVE_INTTYPES_H
printf("seed: %" PRIu32 "\n", seed);
#endif
rng.seed(seed);
boost::unit_test::test_suite* suite = &boost::unit_test::framework::master_test_suite();
suite->p_name.value = "ZlibTest";
uint32_t buf_len = 1024 * 32;
add_tests(suite, gen_uniform_buffer(buf_len, 'a'), buf_len, "uniform");
add_tests(suite, gen_compressible_buffer(buf_len), buf_len, "compressible");
add_tests(suite, gen_random_buffer(buf_len), buf_len, "random");
suite->add(BOOST_TEST_CASE(test_no_write));
return true;
}
float normalvariate(float mean, float sigma)
{
static bool seeded = false;
if (!seeded)
{
// seed generator with #seconds since 1970
rng.seed(static_cast<unsigned> (std::time(0)));
seeded = true;
}
// select desired probability distribution
boost::normal_distribution<float> norm_dist(mean, sigma);
// bind random number generator to distribution, forming a function
boost::variate_generator<boost::mt19937&, boost::normal_distribution<float> > normal_sampler(rng, norm_dist);
// sample from the distribution
return normal_sampler();
}
int main (int argc, char const *argv[]) {
// The default maze size is 20x10. A different size may be specified on
// the command line.
std::size_t x = 20;
std::size_t y = 10;
if (argc == 3) {
x = boost::lexical_cast<std::size_t>(argv[1]);
y = boost::lexical_cast<std::size_t>(argv[2]);
}
random_generator.seed(std::time(0));
maze m(x, y);
random_maze(m);
if (m.solve())
std::cout << "Solved the maze." << std::endl;
else
std::cout << "The maze is not solvable." << std::endl;
std::cout << m << std::endl;
return 0;
}
int main() {
rnd.seed(42);
vertex_iterator vi,vi_end;
Graph m_graph;
auto start = chrono::high_resolution_clock::now();
generate_random_graph(m_graph, 1000/*00*/, 500/*00*/, rnd); // reduced load for Coliru
std::cout << "Generated " << num_vertices(m_graph) << " vertices and " << num_edges(m_graph) << " edges in "
<< chrono::duration_cast<chrono::milliseconds>(chrono::high_resolution_clock::now() - start).count() << "ms\n"
<< "The graph has a cycle? " << std::boolalpha << has_cycle(m_graph) << "\n"
<< "starting selective removal...\n";
start = chrono::high_resolution_clock::now();
size_t count = 0;
for (boost::tie(vi, vi_end) = boost::vertices(m_graph); vi!=vi_end;)
{
if (m_graph[*vi].guid.part1 == 0) {
count++;
clear_vertex(*vi, m_graph);
//std::cout << "." << std::flush;
#if defined(STABLE_IT)
auto toremove = vi++;
boost::remove_vertex(*toremove,m_graph);
#else
boost::remove_vertex(*vi,m_graph);
boost::tie(vi, vi_end) = boost::vertices(m_graph);
#endif
} else
++vi;
}
std::cout << "Done in " << chrono::duration_cast<chrono::milliseconds>(
chrono::high_resolution_clock::now() - start).count() << "ms\n";
std::cout << "After: " << num_vertices(m_graph) << " vertices and " << num_edges(m_graph) << " edges\n";
}
int main(int argc, char ** argv)
{
using namespace std;
using namespace boost;
if(argc <= 3) {
cout << "usage: " << argv[0] << " <grid size> <time> <runs>" << endl;
return -1;
}
int system_random = open("/dev/random", O_RDONLY);
uint32_t seed;
read(system_random, &seed, 4);
close(system_random);
rng.seed(seed);
for(int count = 0; count < atoi(argv[3]);) {
grid_lattice grid(atoi(argv[1]));
while(true) {
double dt = grid.next_event();
grid.time += dt;
if(grid.time > atof(argv[2])) break;
grid.flip();
if(grid.empty()) grid.restart();
}
if(!grid.empty()) {
cout << grid.size() << endl;
// cout << grid << endl;
count++;
}
}
return 0;
}
void RandBipartiteGraph::GenerateGraph()
{
// make n left-vertices
for(unsigned int i = 0; i < n; i++){
AddVariable(i);
}
vector<Vertex*> availableVertexList;
availableVertexList.reserve(n);
gen.seed(time(NULL));
if (c%d != 0){
// make c/d n right-vertices
for(unsigned int i = 0; i < GetRightVerticesNumber(); i++){
availableVertexList.push_back(static_cast<Vertex*>(AddConstraint(i)));
}
}
for (unsigned int i = 0; i < n; i++){
for(unsigned int j = 0; j < c; j++){
AddEdge(static_cast<Vertex*>(variables[i]), GetRandRightVertex(availableVertexList));
}
}
}
void mexFunction( int nlhs, mxArray *plhs[],
int nrhs, const mxArray *prhs[] )
{
//int main(int ac, char* av[]){
// global objects for the program:
/*Cogaps_options_class Cogaps_options(ac, av);
try {
// Cogaps_options_class Cogaps_options(ac, av); // WS -- change from Sasha's
if (Cogaps_options.to_help){
Cogaps_options.help(cout);
return 0;
}
//cout << Cogaps_options;
}
catch( const exception & e) {
cerr <<e.what()<<endl;
return 1;
}
*/
// ===========================================================================
// Initialization of the random number generator.
// Different seeding methods:
// --- fixed seed
//std::vector<unsigned long> ve(2);
//ve[0]=198782;ve[1]=89082;
//boost::random::seed_seq seq(ve);
//rng.seed(seq);
// --- seeded with time
rng.seed(static_cast<boost::uint32_t>(std::time(0)));
//---------------------
// ===========================================================================
// Part 1) Initialization:
// In this section, we read in the system parameters from the paremter file
// parameter.txt, and matrices D and S from datafile.txt.
// Then we initialize A and P in both their atomic domains and
// matrix forms.
// ===========================================================================
//Matlab pointers for inputs
mxArray* DArrayElems;
mxArray* SArrayElems;
mxArray* InputColsPtr;
mxArray* InputRowsPtr;
//Matlab pointers for config
mxArray* nFactorPtr;
mxArray* simulationIDPtr;
mxArray* nEquilPtr;
mxArray* nSamplePtr;
mxArray* QAtomicPtr;
mxArray* alphaAPtr;
mxArray* nIterAPtr;
mxArray* nMaxAPtr;
mxArray* maxGibbsAPtr;
mxArray* lambdaAScalePtr;
mxArray* alphaPPtr;
mxArray* nIterPPtr;
mxArray* nMaxPPtr;
mxArray* maxGibbsPPtr;
mxArray* lambdaPScalePtr;
//Transferring inputs from mex gateway to pointers
DArrayElems = mxGetCell(prhs[0], 0);
SArrayElems = mxGetCell(prhs[0], 1);
InputColsPtr = mxGetCell(prhs[0], 2);
InputRowsPtr = mxGetCell(prhs[0], 3);
nFactorPtr = mxGetCell(prhs[0], 4);
simulationIDPtr = mxGetCell(prhs[0], 5);
nEquilPtr = mxGetCell(prhs[0], 6);
nSamplePtr = mxGetCell(prhs[0], 7);
QAtomicPtr = mxGetCell(prhs[0], 8);
alphaAPtr = mxGetCell(prhs[0], 9);
nIterAPtr = mxGetCell(prhs[0], 10);
nMaxAPtr = mxGetCell(prhs[0], 11);
maxGibbsAPtr = mxGetCell(prhs[0], 12);
lambdaAScalePtr = mxGetCell(prhs[0], 13);
alphaPPtr = mxGetCell(prhs[0], 14);
nIterPPtr = mxGetCell(prhs[0], 15);
nMaxPPtr = mxGetCell(prhs[0], 16);
maxGibbsPPtr = mxGetCell(prhs[0], 17);
lambdaPScalePtr = mxGetCell(prhs[0], 18);
//creating native C style data types
double* DArray;
double* SArray;
double* numInputRows;
double* numInputCols;
//Establish the C types for the Config Information
double* nFactorC = mxGetPr(nFactorPtr);
int ptrLength;
char* simulationIDC;
ptrLength = mxGetNumberOfElements(simulationIDPtr) + 1;
simulationIDC = (char *)(mxCalloc(ptrLength, sizeof(char)));
mxGetString(simulationIDPtr, simulationIDC, ptrLength);
double* nEquilC = mxGetPr(nEquilPtr);
double* nSampleC = mxGetPr(nSamplePtr);
double* QAtomicC = mxGetPr(QAtomicPtr);
double* alphaAC = mxGetPr(alphaAPtr);
double* nIterAC = mxGetPr(nIterAPtr);
double* nMaxAC = mxGetPr(nMaxAPtr);
double* maxGibbsAC = mxGetPr(maxGibbsAPtr);
double* lambdaAScaleC = mxGetPr(lambdaAScalePtr);
double* alphaPC = mxGetPr(alphaPPtr);
double* nIterPC = mxGetPr(nIterPPtr);
double* nMaxPC = mxGetPr(nMaxPPtr);
double* maxGibbsPC = mxGetPr(maxGibbsPPtr);
double* lambdaPScaleC = mxGetPr(lambdaPScalePtr);
int nIRows;
int nICols;
//converting from pointer to matlab type then to C array
DArray = mxGetPr(DArrayElems);
SArray = mxGetPr(SArrayElems);
numInputCols = mxGetPr(InputRowsPtr);
numInputRows = mxGetPr(InputColsPtr);
//removing need for arrays for dimensions
nIRows = numInputRows[0];
nICols = numInputCols[0];
//Putting C style arrays into C++ vectors to pass to the Gibbs Sampler
vector< vector<double> > DVector(nIRows, vector<double>(nICols));
vector< vector<double> > SVector(nIRows, vector<double>(nICols));
for(int i = 0; i < nIRows; i++)
{
for(int j = 0; j < nICols; j++)
{
DVector[i][j] = DArray[i + (j*nIRows)];
}
}
for(int i = 0; i < nIRows; i++)
{
for(int j = 0; j < nICols; j++)
{
SVector[i][j] = SArray[i + (j*nIRows)];
}
}
/*Testing output of matrices
cout << endl;
for(int i = 0; i < nIRows; i++)
{
for(int j = 0; j < nICols; j++)
{
cout << DVector[i][j] << " ";
}
cout << endl;
}
cout << endl;
for(int i = 0; i < nIRows; i++)
{
for(int j = 0; j < nICols; j++)
{
cout << SVector[i][j] << " ";
}
cout << endl;
}
*/
// Parameters or data to be read in:
unsigned long nEquil = nEquilC[0]; //Cogaps_options.nEquil; // # outer loop iterations
// for equilibration
unsigned long nSample = nSampleC[0];//Cogaps_options.nSample; // # outer loop iterations
// for sampling
unsigned int nFactor = nFactorC[0]; //Cogaps_options.nFactor; // # patterns
double alphaA = alphaAC[0]; //Cogaps_options.alphaA;
double alphaP = alphaPC[0]; //Cogaps_options.alphaP;
double nMaxA = nMaxAC[0];//Cogaps_options.nMaxA; // max. number of atoms in A
double nMaxP = nMaxPC[0];//Cogaps_options.nMaxP; // number of atomic bins for P
//string datafile = "Data2.txt";//Cogaps_options.datafile; // File for D
//string variancefile = "Noise2.txt";//Cogaps_options.variancefile; // File for S
string simulation_id = simulationIDC;//Cogaps_options.simulation_id; // simulation id
unsigned long nIterA = nIterAC[0];//Cogaps_options.nIterA; // initial # of inner-loop iterations for A
unsigned long nIterP = nIterPC[0];//Cogaps_options.nIterP; // initial # of inner-loop iterations for P
double max_gibbsmass_paraA = maxGibbsAC[0];//Cogaps_options.max_gibbsmass_paraA;
// maximum gibbs mass parameter for A
double max_gibbsmass_paraP = maxGibbsPC[0];//Cogaps_options.max_gibbsmass_paraP;
// maximum gibbs mass parameter for P
double lambdaA_scale_factor = lambdaAScaleC[0];//Cogaps_options.lambdaA_scale_factor;
// scale factor for lambdaA
double lambdaP_scale_factor = lambdaPScaleC[0];//Cogaps_options.lambdaP_scale_factor;
// scale factor for lambdaP
bool Q_output_atomic = QAtomicC[0];//Cogaps_options.Q_output_atomic;
// whether to output the atomic space info
// Parameters or structures to be calculated or constructed:
unsigned int nRow; // number of items in observation (= # of genes)
unsigned int nCol; // number of observation (= # of arrays)
unsigned int nBinsA; // number of atomic bins for A
unsigned int nBinsP; // number of atomic bins for P
double lambdaA;
double lambdaP;
//unsigned long nIterA; // number of inner loop iterations for A
//unsigned long nIterP; // number of inner loop iterations for P
//atomic At, Pt; // atomic space for A and P respectively
unsigned long atomicSize; // number of atomic points
char label_A = 'A'; // label for matrix A
char label_P = 'P'; // label for matrix P
char label_D = 'D'; // label for matrix D
char label_S = 'S';// label for matrix S
// Output parameters and computing info to files:
/* Commented out for Matlab and R Versions
char outputFilename[80];
strcpy(outputFilename,simulation_id.c_str());
strcat(outputFilename,"_computing_info.txt");
ofstream outputFile;
outputFile.open(outputFilename,ios::out); // start by deleting previous content and
// rewriting the file
outputFile << "Common parameters and info:" << endl;
outputFile << "input data file: " << datafile << endl;
outputFile << "input variance file: " << variancefile << endl;
outputFile << "simulation id: " << simulation_id << endl;
outputFile << "nFactor = " << nFactor << endl;
outputFile << "nEquil = " << nEquil << endl;
outputFile << "nSample = " << nSample << endl;
outputFile << "Q_output_atomic (bool) = " << Q_output_atomic << endl << endl;
outputFile << "Parameters for A:" << endl;
outputFile << "alphaA = " << alphaA << endl;
outputFile << "nMaxA = " << nMaxA << endl;
outputFile << "nIterA = " << nIterA << endl;
outputFile << "max_gibbsmass_paraA = " << max_gibbsmass_paraA << endl;
outputFile << "lambdaA_scale_factor = " << lambdaA_scale_factor << endl << endl;
outputFile << "Parameters for P:" << endl;
outputFile << "alphaP = " << alphaP << endl;
outputFile << "nMaxP = " << nMaxP << endl;
outputFile << "nIterP = " << nIterP << endl;
outputFile << "max_gibbsmass_paraP = " << max_gibbsmass_paraP << endl;
outputFile << "lambdaP_scale_factor = " << lambdaP_scale_factor << endl << endl;
outputFile.close();
*/
// ---------------------------------------------------------------------------
// Initialize the GibbsSampler.
/*Regular Version
GibbsSampler GibbsSamp(nEquil,nSample,nFactor, // construct GibbsSampler and
alphaA,alphaP,nMaxA,nMaxP,// Read in D and S matrices
nIterA,nIterP,
max_gibbsmass_paraA, max_gibbsmass_paraP,
lambdaA_scale_factor, lambdaP_scale_factor,
atomicSize,
label_A,label_P,label_D,label_S,
datafile,variancefile,simulation_id);
*/
//R and Matlab Version
GibbsSampler GibbsSamp(nEquil,nSample,nFactor, // construct GibbsSampler and
alphaA,alphaP,nMaxA,nMaxP,// Read in D and S matrices
nIterA,nIterP,
max_gibbsmass_paraA, max_gibbsmass_paraP,
lambdaA_scale_factor, lambdaP_scale_factor,
atomicSize,
label_A,label_P,label_D,label_S,
DVector,SVector,simulation_id);
// ---------------------------------------------------------------------------
// Based on the information of D, construct and initialize for A and P both
// the matrices and atomic spaces.
GibbsSamp.init_AMatrix_and_PMatrix(); // initialize A and P matrices
GibbsSamp.init_AAtomicdomain_and_PAtomicdomain(); // intialize atomic spaces
// A and P
GibbsSamp.init_sysChi2(); // initialize the system chi2 value
// ===========================================================================
// Part 2) Equilibration:
// In this section, we let the system eqilibrate with nEquil outer loop
// iterations. Within each outer loop iteration, A is iterated nIterA times
// and P is iterated nIterP times. After equilibration, we update nIterA and
// nIterP according to the expected number of atoms in the atomic spaces
// of A and P respectively.
// ===========================================================================
// --------- temp for initializing output to chi2.txt
/*
char outputchi2_Filename[80];
strcpy(outputchi2_Filename,simulation_id.c_str());
strcat(outputchi2_Filename,"_chi2.txt");
ofstream outputchi2_File;
outputchi2_File.open(outputchi2_Filename,ios::out);
outputchi2_File << "chi2" << endl;
outputchi2_File.close();
*/
// --------------
double chi2;
//Matlab control variables
double tempChiSq;
double tempAtomA;
double tempAtomP;
int outCount = 0;
int numOutputs = 100;
int totalChiSize = nSample + nEquil;
//Establishing Matlab containers for atoms and chisquare
double* nAEquil;
mxArray* nAEMx;
nAEMx = mxCreateDoubleMatrix(1, nEquil, mxREAL);
nAEquil = mxGetPr(nAEMx);
double* nASamp;
mxArray* nASMx;
nASMx = mxCreateDoubleMatrix(1, nSample, mxREAL);
nASamp = mxGetPr(nASMx);
double* nPEquil;
mxArray* nPEMx;
nPEMx = mxCreateDoubleMatrix(1, nEquil, mxREAL);
nPEquil = mxGetPr(nPEMx);
double* nPSamp;
mxArray* nPSMx;
nPSMx = mxCreateDoubleMatrix(1, nSample, mxREAL);
nPSamp = mxGetPr(nPSMx);
double* chiVect;
mxArray* chiMx;
chiMx = mxCreateDoubleMatrix(1, totalChiSize, mxREAL);
chiVect = mxGetPr(chiMx);
for (unsigned long ext_iter=1; ext_iter <= nEquil; ++ext_iter){
GibbsSamp.set_iter(ext_iter);
GibbsSamp.set_AnnealingTemperature();
for (unsigned long iterA=1; iterA <= nIterA; ++iterA){
GibbsSamp.update('A');
GibbsSamp.check_atomic_matrix_consistency('A');
//GibbsSamp.check_atomic_matrix_consistency('A');
//GibbsSamp.detail_check(outputchi2_Filename);
}
for (unsigned long iterP=1; iterP <= nIterP; ++iterP){
GibbsSamp.update('P');
GibbsSamp.check_atomic_matrix_consistency('P');
//GibbsSamp.check_atomic_matrix_consistency('P');
//GibbsSamp.detail_check(outputchi2_Filename);
}
tempChiSq = GibbsSamp.get_sysChi2();
chiVect[(ext_iter)-1] = tempChiSq;
tempAtomA = GibbsSamp.getTotNumAtoms('A');
tempAtomP = GibbsSamp.getTotNumAtoms('P');
nAEquil[outCount] = tempAtomA;
nPEquil[outCount] = tempAtomP;
outCount++;
// ----------- output computing info ---------
if ( ext_iter % numOutputs == 0){
//chi2 = 2.*GibbsSamp.cal_logLikelihood();
cout << "Equil: " << ext_iter << " of " << nEquil <<
" ,nA: " << tempAtomA <<
" ,nP: " << tempAtomP <<
// " ,chi2 = " << chi2 <<
" ,System Chi2 = " << tempChiSq << endl;
}
// -------------------------------------------
// re-calculate nIterA and nIterP to the expected number of atoms
nIterA = (unsigned long) randgen('P',max((double) GibbsSamp.getTotNumAtoms('A'),10.));
nIterP = (unsigned long) randgen('P',max((double) GibbsSamp.getTotNumAtoms('P'),10.));
//nIterA = (unsigned long) randgen('P',(double) GibbsSamp.getTotNumAtoms('A')+10.);
//nIterP = (unsigned long) randgen('P',(double) GibbsSamp.getTotNumAtoms('P')+10.);
// --------------------------------------------
} // end of for-block for equilibration
// ===========================================================================
// Part 3) Sampling:
// After the system equilibriates in Part 2, we sample the systems with an
// outer loop of nSample iterations. Within each outer loop iteration, A is
// iterated nIterA times and P is iterated nIterP times. After sampling,
// we update nIterA and nIterP according to the expected number of atoms in
// the atomic spaces of A and P respectively.
// ===========================================================================
unsigned int statindx = 0;
outCount = 0;
for (unsigned long ext_iter=1; ext_iter <= nSample; ++ext_iter){
for (unsigned long iterA=1; iterA <= nIterA; ++iterA){
GibbsSamp.update('A');
//GibbsSamp.check_atomic_matrix_consistency('A');
//GibbsSamp.detail_check(outputchi2_Filename);
}
GibbsSamp.check_atomic_matrix_consistency('A');
for (unsigned long iterP=1; iterP <= nIterP; ++iterP){
GibbsSamp.update('P');
//GibbsSamp.check_atomic_matrix_consistency('P');
//GibbsSamp.detail_check(outputchi2_Filename);
}
GibbsSamp.check_atomic_matrix_consistency('P');
if (Q_output_atomic == true){
GibbsSamp.output_atomicdomain('A',ext_iter);
GibbsSamp.output_atomicdomain('P',ext_iter);
}
statindx += 1;
GibbsSamp.compute_statistics_prepare_matrices(statindx);
tempChiSq = GibbsSamp.get_sysChi2();
chiVect[(nEquil + ext_iter)-1] = tempChiSq;
tempAtomA = GibbsSamp.getTotNumAtoms('A');
tempAtomP = GibbsSamp.getTotNumAtoms('P');
nASamp[outCount] = tempAtomA;
nPSamp[outCount] = tempAtomP;
outCount++;
// ----------- output computing info ---------
if ( ext_iter % numOutputs == 0){
// chi2 = 2.*GibbsSamp.cal_logLikelihood();
//GibbsSamp.output_atomicdomain('A',(unsigned long) statindx);
//GibbsSamp.output_atomicdomain('P',(unsigned long) statindx);
cout << "Samp: " << ext_iter << " of " << nSample <<
" ,nA: " << tempAtomA <<
" ,nP: " << tempAtomP <<
// " ,chi2 = " << chi2 <<
" , System Chi2 = " << tempChiSq << endl;
if (ext_iter == nSample){
chi2 = 2.*GibbsSamp.cal_logLikelihood();
cout << " *** Check value of final chi2: " << chi2 << " **** " << endl;
}
}
// -------------------------------------------
// re-calculate nIterA and nIterP to the expected number of atoms
nIterA = (unsigned long) randgen('P',max((double) GibbsSamp.getTotNumAtoms('A'),10.));
nIterP = (unsigned long) randgen('P',max((double) GibbsSamp.getTotNumAtoms('P'),10.));
//nIterA = (unsigned long) randgen('P',(double) GibbsSamp.getTotNumAtoms('A')+10.);
//nIterP = (unsigned long) randgen('P',(double) GibbsSamp.getTotNumAtoms('P')+10.);
// --------------------------------------------
} // end of for-block for Sampling
// ===========================================================================
// Part 4) Calculate statistics:
// In this final section, we calculate all statistics pertaining to the final
// sample and check the results.
// ===========================================================================
char outputAmean_Filename[80];
strcpy(outputAmean_Filename,simulation_id.c_str());
strcat(outputAmean_Filename,"_Amean.txt");
char outputAsd_Filename[80];
strcpy(outputAsd_Filename,simulation_id.c_str());
strcat(outputAsd_Filename,"_Asd.txt");
char outputPmean_Filename[80];
strcpy(outputPmean_Filename,simulation_id.c_str());
strcat(outputPmean_Filename,"_Pmean.txt");
char outputPsd_Filename[80];
strcpy(outputPsd_Filename,simulation_id.c_str());
strcat(outputPsd_Filename,"_Psd.txt");
char outputAPmean_Filename[80];
strcpy(outputAPmean_Filename,simulation_id.c_str());
strcat(outputAPmean_Filename,"_APmean.txt");
/* GibbsSamp.compute_statistics(outputFilename,
outputAmean_Filename,outputAsd_Filename,
outputPmean_Filename,outputPsd_Filename,
outputAPmean_Filename,
statindx); // compute statistics like mean and s.d.n */
vector<vector <double> > AMeanVector;
vector<vector <double> > AStdVector;
vector<vector <double> > PMeanVector;
vector<vector <double> > PStdVector;
GibbsSamp.compute_statistics(statindx,
AMeanVector, AStdVector, PMeanVector, PStdVector);
/*
cout << endl;
for(int i = 0; i < AMeanVector.size(); i++)
{
for(int j = 0; j < AMeanVector[0].size(); j++)
{
cout << AMeanVector[i][j] << " ";
}
cout << endl;
}
*/
/*double *x,*y;
size_t mrows,ncols;
mrows = mxGetM(prhs[0]);
ncols = mxGetN(prhs[0]);
plhs[0] = mxCreateDoubleMatrix((mwSize)mrows, (mwSize)ncols, mxREAL);
x = mxGetPr(prhs[0]);
y = mxGetPr(plhs[0]);*/
//Get Dims of Matrices for matlab
int nrowA, ncolA;
int nrowP, ncolP;
ncolA = AMeanVector.size();
nrowA = AMeanVector[0].size();
ncolP = PMeanVector.size();
nrowP = PMeanVector[0].size();
//Declare Matrices for matlab
double* AMeanMatArray;
double* AStdMatArray;
double* PMeanMatArray;
double* PStdMatArray;
//Convertible Matlab datatypes
mxArray* AMeanMx;
mxArray* AStdMx;
mxArray* PMeanMx;
mxArray* PStdMx;
//Allocate memory for the Matlab Matrices and establish their data types
AMeanMx = mxCreateDoubleMatrix(ncolA, nrowA, mxREAL);
AStdMx = mxCreateDoubleMatrix(ncolA, nrowA, mxREAL);
PMeanMx = mxCreateDoubleMatrix(ncolP, nrowP, mxREAL);
PStdMx = mxCreateDoubleMatrix(ncolP, nrowP, mxREAL);
//Steps to Create the Cell Array to pass back the data matrices
mxArray* TempArry;
const int* dims;
TempArry = mxCreateDoubleMatrix(9, 9, mxREAL); //TODO, currently creating the dimensions of a cell Matrix by creating a temp matrix due to lack of understanding of mex datatypes, fix!
dims = mxGetDimensions(TempArry);
plhs[0] = mxCreateCellArray(1, dims);
//Fill the matrices individually
AMeanMatArray = mxGetPr(AMeanMx);
double tempVectValue;
for(int i = 0; i < ncolA; i++)
{
for(int j = 0; j < nrowA; j++)
{
tempVectValue = AMeanVector[i][j];
AMeanMatArray[i + (j*ncolA)] = tempVectValue;
}
}
AStdMatArray = mxGetPr(AStdMx);
for(int i = 0; i < ncolA; i++)
{
for(int j = 0; j < nrowA; j++)
{
tempVectValue = AStdVector[i][j];
AStdMatArray[i + (j*ncolA)] = tempVectValue;
}
}
PMeanMatArray = mxGetPr(PMeanMx);
for(int i = 0; i < ncolP; i++)
{
for(int j = 0; j < nrowP; j++)
{
tempVectValue = PMeanVector[i][j];
PMeanMatArray[i + (j*ncolP)] = tempVectValue;
}
}
PStdMatArray = mxGetPr(PStdMx);
for(int i = 0; i < ncolP; i++)
{
for(int j = 0; j < nrowP; j++)
{
tempVectValue = PStdVector[i][j];
PStdMatArray[i + (j*ncolP)] = tempVectValue;
}
}
mxSetCell(plhs[0], 0, AMeanMx);
mxSetCell(plhs[0], 1, AStdMx);
mxSetCell(plhs[0], 2, PMeanMx);
mxSetCell(plhs[0], 3, PStdMx);
mxSetCell(plhs[0], 4, nAEMx);
mxSetCell(plhs[0], 5, nASMx);
mxSetCell(plhs[0], 6, nPEMx);
mxSetCell(plhs[0], 7, nPSMx);
mxSetCell(plhs[0], 8, chiMx);
}
int
main(int argc, char *argv[])
{
// Parse and read the command line arguments.
if (argc != 5)
{
cerr << "Usage: " << PNAME
<< " <min-weight> <max-weight> <#vertices> <#tests>\n";
exit(1);
}
long int min_weight = strtol(argv[1], 0, 0);
long int max_weight = strtol(argv[2], 0, 0);
long int n_vertices = strtol(argv[3], 0, 0);
long int n_tests = strtol(argv[4], 0, 0);
if (errno)
{
cerr << "Bad arguments\n";
exit(1);
}
// Check consistency of command line arguments.
if (n_tests < 0)
{
cerr << "the number of tests must be a positive integer\n";
exit(1);
}
if (n_vertices > numeric_limits<unsigned>::digits)
{
cerr << "Too many vertices\n"
<< "Maximum n.o. vertices is "
<< numeric_limits<unsigned>::digits << "\n";
exit(1);
}
if (n_vertices < 2)
{
cerr << "Too few vertices\n"
<< "must be at least two\n";
exit(1);
}
if (min_weight > max_weight)
{
cerr << "min weight must be greater than max weight\n";
exit(1);
}
if (abs(max_weight) >= numeric_limits<int>::max() / n_vertices ||
abs(min_weight) >= numeric_limits<int>::max() / n_vertices)
{
cerr << "weights too large\n";
exit(1);
}
// Initialize the random number generator.
timeval tv;
gettimeofday(&tv, 0);
g_generator.seed(tv.tv_sec * 1000000 + tv.tv_usec);
boost::uniform_int<int> uniform_int_dist(min_weight, max_weight);
boost::variate_generator<mt19937, uniform_int<int> > rng_i(g_generator, uniform_int_dist);
// Perform the tests.
enum {MAX_SPAN_0, MAX_SPAN_1, MAX_SPAN_2,
MAX_NOSPAN_0, MAX_NOSPAN_1, MAX_NOSPAN_2,
MIN_SPAN_0, MIN_SPAN_1, MIN_SPAN_2,
MIN_NOSPAN_0, MIN_NOSPAN_1, MIN_NOSPAN_2,
N_CASES};
string case_names[] = {
"MAX_SPAN_0", "MAX_SPAN_1", "MAX_SPAN_2",
"MAX_NOSPAN_0", "MAX_NOSPAN_1", "MAX_NOSPAN_2",
"MIN_SPAN_0", "MIN_SPAN_1", "MIN_SPAN_2",
"MIN_NOSPAN_0", "MIN_NOSPAN_1", "MIN_NOSPAN_2"
};
complete_graph g(n_vertices);
multi_array<int, 2> weights(extents[n_vertices][n_vertices]);
vector<Vertex> parent(n_vertices);
Vertex roots[] = {0, 1};
vector<Edge> branching;
vector< vector< Edge > > all_branchings(N_CASES);
vector<int> ans(N_CASES);
while (n_tests--)
{
// Clear all the branchings from previous iteration.
BOOST_FOREACH (vector<Edge> &vec, all_branchings)
{
vec.clear();
}
// create new weights
int *end = weights.origin() + n_vertices * n_vertices;
for (int *ip = weights.origin(); ip != end; ++ip)
{
*ip = rng_i();
}
// run edmonds algorithm for a few cases. The cases will be
// the cross product of the following properties:
// optimum-is-maximum x attempt-to-span x num-specified-roots
// where num-specified roots is either 0, 1, or 2. Also the
// specified roots are either none, the vertex 0, or the
// vertices 0 and 1.
edmonds_optimum_branching<true, true, true>
(g, identity_property_map(), weights.origin(),
roots, roots + 0, back_inserter(all_branchings[MAX_SPAN_0]));
edmonds_optimum_branching<true, true, true>
(g, identity_property_map(), weights.origin(),
roots, roots + 1, back_inserter(all_branchings[MAX_SPAN_1]));
edmonds_optimum_branching<true, true, true>
(g, identity_property_map(), weights.origin(),
roots, roots + 2, back_inserter(all_branchings[MAX_SPAN_2]));
edmonds_optimum_branching<true, false, true>
(g, identity_property_map(), weights.origin(),
roots, roots + 0, back_inserter(all_branchings[MAX_NOSPAN_0]));
edmonds_optimum_branching<true, false, true>
(g, identity_property_map(), weights.origin(),
roots, roots + 1, back_inserter(all_branchings[MAX_NOSPAN_1]));
edmonds_optimum_branching<true, false, true>
(g, identity_property_map(), weights.origin(),
roots, roots + 2, back_inserter(all_branchings[MAX_NOSPAN_2]));
edmonds_optimum_branching<false, true, true>
(g, identity_property_map(), weights.origin(),
roots, roots + 0, back_inserter(all_branchings[MIN_SPAN_0]));
edmonds_optimum_branching<false, true, true>
(g, identity_property_map(), weights.origin(),
roots, roots + 1, back_inserter(all_branchings[MIN_SPAN_1]));
edmonds_optimum_branching<false, true, true>
(g, identity_property_map(), weights.origin(),
roots, roots + 2, back_inserter(all_branchings[MIN_SPAN_2]));
edmonds_optimum_branching<false, false, true>
(g, identity_property_map(), weights.origin(),
roots, roots + 0, back_inserter(all_branchings[MIN_NOSPAN_0]));
edmonds_optimum_branching<false, false, true>
(g, identity_property_map(), weights.origin(),
roots, roots + 1, back_inserter(all_branchings[MIN_NOSPAN_1]));
edmonds_optimum_branching<false, false, true>
(g, identity_property_map(), weights.origin(),
roots, roots + 2, back_inserter(all_branchings[MIN_NOSPAN_2]));
for (int i = 0; i < N_CASES; ++i)
{
ans[i] = 0;
BOOST_FOREACH (Edge e, all_branchings[i])
{
ans[i] += weights[source(e, g)][target(e, g)];
}
}
// Exhaustively test all branchings and check if the answers
// from edmonds's algorithm is optimal.
int bad_case = -1;
fill(parent.begin(), parent.end(), 0);
do
{
if (is_cyclic(parent))
continue;
int weight = 0;
for (int i = 0; i < n_vertices; ++i)
{
if (parent[i] == i)
continue;
weight += weights[parent[i]][i];
}
int num_roots = 0;
for (int i = 0; i < n_vertices; ++i)
{
num_roots += parent[i] == i ? 1 : 0;
}
bad_case = MAX_NOSPAN_0;
if (weight > ans[bad_case])
break;
bad_case = MIN_NOSPAN_0;
if (weight < ans[bad_case])
break;
bad_case = MAX_NOSPAN_1;
if (parent[0] == 0 && weight > ans[bad_case])
break;
bad_case = MIN_NOSPAN_1;
if (parent[0] == 0 && weight < ans[bad_case])
break;
bad_case = MAX_NOSPAN_2;
if (parent[0] == 0 && parent[1] == 1 && weight > ans[bad_case])
break;
bad_case = MIN_NOSPAN_2;
if (parent[0] == 0 && parent[1] == 1 && weight < ans[bad_case])
break;
bad_case = MAX_SPAN_0;
if (num_roots == 1 && weight > ans[bad_case])
break;
bad_case = MIN_SPAN_0;
if (num_roots == 1 && weight < ans[bad_case])
break;
bad_case = MAX_SPAN_1;
if (num_roots == 1 && parent[0] == 0 && weight > ans[bad_case])
break;
bad_case = MIN_SPAN_1;
if (num_roots == 1 && parent[0] == 0 && weight < ans[bad_case])
break;
bad_case = MAX_SPAN_2;
if (num_roots == 1 && parent[0] == 0 && parent[1] == 1 &&
weight > ans[bad_case])
break;
bad_case = MIN_SPAN_2;
if (num_roots == 1 && parent[0] == 0 && parent[1] == 1 &&
weight < ans[bad_case])
break;
bad_case = -1;
} while (next_parent(parent));
// If we broke out of the loop prematurely, then something was
// wrong.
if (bad_case > 0)
{
cout << "\ntest failed.\n";
cout << "weights:\n";
for (int i = 0; i < n_vertices; ++i)
{
for (int j = 0; j < n_vertices; ++j)
{
cout << weights[i][j] << " ";
}
cout << "\n";
}
cout << "\n";
int weight = 0;
for (int i = 0; i < n_vertices; ++i)
{
if (parent[i] == i)
continue;
weight += weights[parent[i]][i];
}
cout << "counter example (weight = " << weight << "):\n";
for (int i = 0; i < n_vertices; ++i)
{
if (parent[i] == i)
continue;
cout << "(" << parent[i] << ", " << i << ")\t";
}
cout << "\n";
weight = 0;
cout << "case: " << case_names[bad_case] << ":\n";
BOOST_FOREACH (Edge e, all_branchings[bad_case])
{
cout << "(" << source(e, g) << ", " << target(e, g) << ")\t";
weight += weights[source(e, g)][target(e, g)];
}
cout << "\n";
cout << "weight:\t" << weight << "\n";
exit(1);
}
servers_chooser()
{
boost::random_device dev;
rng_.seed(dev());
}
void initrand(unsigned int seed) {
rng.seed(seed);
}
int main (int argc, char* argv[])
{
cout << endl;
cout << " MAP AND MARGINAL INFERENCE EXAMPLE";
cout << endl << endl;
/*------------------------------------------------------------------------------
*
* PREPARE TRAINING DATA
*
*----------------------------------------------------------------------------*/
//
// Generate the types of nodes and edges
//
size_t N_classes_type_1 = 11;
size_t N_nf = 5;
size_t N_ef = 4;
// OBJECTS
CNodeTypePtr simpleNodeType1Ptr( new CNodeType(N_classes_type_1,
N_nf,
string("Objects") ) );
CEdgeTypePtr simpleEdgeType1Ptr ( new CEdgeType(N_ef,
simpleNodeType1Ptr,
simpleNodeType1Ptr,
string("Edges between two objects")) );
trainingDataset.addNodeType( simpleNodeType1Ptr );
trainingDataset.addEdgeType( simpleEdgeType1Ptr );
//
// Create a set of training graphs
//
vector<CGraph> graphs;
vector<std::map<size_t,size_t> > groundTruths;
/*--------------------------------- Graph 1 ----------------------------------*/
cout << "Graph 1" << endl;
CGraph graph1;
std::map<size_t,size_t> groundTruth1;
Eigen::MatrixXd g1_nf(11,5);
g1_nf << 1, 0.725256, 0.347833, 2.32114, 1,
0, 0, 1.46145, 2.86637, 1,
1, 0.513579, 0.398025, 1.85308, 1,
0, 0.896507, 0.969337, 1.7035, 1,
0, 0.748331, 0.585186, 1.47532, 1,
1, 0.963636, 0.125714, 1, 1,
1, 0.93444, 0.212592, 1.52702, 1,
1, 1.05757, 0.236803, 3.22358, 1,
0, 0.398408, 0.131399, 1, 1,
0, 0.0300602, 0.129521, 2.03183, 1,
1, 0.469344, 0.505431, 1.45506, 1;
Eigen::MatrixXi g1_adj(11,11);
g1_adj << 0, 1, 1, 1, 1, 1, 0, 0, 1, 1, 0,
1, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0,
1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1, 0, 0, 0, 1, 1, 1, 1, 1, 0, 1,
1, 0, 0, 1, 0, 1, 1, 1, 1, 1, 1,
1, 0, 0, 1, 1, 0, 1, 0, 0, 0, 0,
0, 0, 0, 1, 1, 1, 0, 1, 1, 0, 1,
0, 0, 0, 1, 1, 0, 1, 0, 0, 0, 1,
1, 1, 0, 1, 1, 0, 1, 0, 0, 1, 0,
1, 1, 0, 0, 1, 0, 0, 0, 1, 0, 0,
0, 0, 0, 1, 1, 0, 1, 1, 0, 0, 0;
Eigen::MatrixXi isOn_relation(11,11);
isOn_relation.setZero();
isOn_relation(3,5) = 1;
isOn_relation(4,5) = 1;
Eigen::MatrixXi coplanar_relation(11,11);
coplanar_relation.setZero();
vector<Eigen::MatrixXi> g1_relations;
g1_relations.push_back( isOn_relation );
g1_relations.push_back( coplanar_relation );
Eigen::VectorXi g1_groundTruth(11);
g1_groundTruth << 4, 0, 1, 2, 2, 6, 1, 1, 5, 0, 3;
fillGraph( graph1, g1_nf, simpleNodeType1Ptr,
g1_adj, simpleEdgeType1Ptr, g1_relations,
g1_groundTruth, groundTruth1 );
graphs.push_back( graph1 );
groundTruths.push_back( groundTruth1 );
cout << graph1 << endl;
/*--------------------------------- Graph 2 ----------------------------------*/
cout << "Graph 2" << endl;
CGraph graph2;
std::map<size_t,size_t> groundTruth2;
Eigen::MatrixXd g2_nf(5,5);
g2_nf << 0, 0.745944, 1.20098, 1.68726, 1,
1, 0.523566, 0.255477, 1.72744, 1,
0, 0, 1.49268, 2.54707, 1,
1, 0.359525, 0.268375, 2.8157, 1,
0, 0.446692, 0.129895, 1, 1;
Eigen::MatrixXi g2_adj(5,5);
g2_adj << 0, 1, 0, 1, 1,
1, 0, 1, 1, 0,
0, 1, 0, 1, 1,
1, 1, 1, 0, 1,
1, 0, 1, 1, 0;
isOn_relation.resize(5,5);
isOn_relation.setZero();
coplanar_relation.resize(5,5);
coplanar_relation.setZero();
vector<Eigen::MatrixXi> g2_relations;
g2_relations.push_back( isOn_relation );
g2_relations.push_back( coplanar_relation );
Eigen::VectorXi g2_groundTruth(5);
g2_groundTruth << 2, 3, 0, 3, 5;
fillGraph( graph2, g2_nf, simpleNodeType1Ptr,
g2_adj, simpleEdgeType1Ptr, g2_relations,
g2_groundTruth, groundTruth2 );
graphs.push_back( graph2 );
groundTruths.push_back( groundTruth2 );
///*--------------------------------- Graph 3 ----------------------------------*/
cout << "Graph 3" << endl;
CGraph graph3;
std::map<size_t,size_t> groundTruth3;
Eigen::MatrixXd g3_nf(10,5);
g3_nf << 1, 1.56711, 3.42284, 2.00889, 1,
1, 1.02852, 0.339804, 1.54119, 1,
1, 1.02507, 0.317727, 1.77383, 1,
0, 0.880216, 1.45163, 2.04593, 1,
1, 1.04219, 0.229974, 1.77708, 1,
1, 1.66558, 0.171681, 2.3654, 1,
0, 0.488131, 0.233171, 1.51503, 1,
1, 0.699061, 0.120062, 1.55154, 1,
0, 0, 2.84036, 1.99605, 1,
1, 0.167627, 0.352816, 3.80501, 1;
Eigen::MatrixXi g3_adj(10,10);
g3_adj << 0, 1, 1, 1, 1, 1, 0, 0, 0, 1,
1, 0, 1, 1, 0, 0, 0, 0, 0, 0,
1, 1, 0, 1, 1, 1, 1, 1, 0, 1,
1, 1, 1, 0, 1, 1, 1, 1, 0, 1,
1, 0, 1, 1, 0, 1, 1, 1, 0, 1,
1, 0, 1, 1, 1, 0, 0, 0, 0, 0,
0, 0, 1, 1, 1, 0, 0, 1, 1, 0,
0, 0, 1, 1, 1, 0, 1, 0, 1, 0,
0, 0, 0, 0, 0, 0, 1, 1, 0, 1,
1, 0, 1, 1, 1, 0, 0, 0, 1, 0;
isOn_relation.resize(10,10);
isOn_relation.setZero();
isOn_relation(1,3) = 1;
isOn_relation(2,3) = 1;
isOn_relation(3,4) = 1;
coplanar_relation.resize(10,10);
coplanar_relation.setZero();
vector<Eigen::MatrixXi> g3_relations;
g3_relations.push_back( isOn_relation );
g3_relations.push_back( coplanar_relation );
Eigen::VectorXi g3_groundTruth(10);
g3_groundTruth << 1, 6, 6, 2, 9, 10, 5, 4, 0, 1;
fillGraph( graph3, g3_nf, simpleNodeType1Ptr,
g3_adj, simpleEdgeType1Ptr, g3_relations,
g3_groundTruth, groundTruth3 );
graphs.push_back( graph3 );
groundTruths.push_back( groundTruth3 );
///*--------------------------------- Graph 4 ----------------------------------*/
cout << "Graph 4" << endl;
CGraph graph4;
std::map<size_t,size_t> groundTruth4;
Eigen::MatrixXd g4_nf(7,5);
g4_nf << 1, 1.58673, 2.5301, 1.79962, 1,
1, 1.01892, 0.198802, 1.46393, 1,
1, 1.00455, 0.230172, 1.67413, 1,
0, 0.744138, 0.89911, 1.57856, 1,
0, 0.477457, 0.187906, 1.87622, 1,
1, 0.7026, 0.113952, 1.44738, 1,
0, 0, 2.75497, 1.87592, 1;
Eigen::MatrixXi g4_adj(7,7);
g4_adj << 0, 1, 1, 1, 0, 0, 0,
1, 0, 1, 1, 1, 1, 0,
1, 1, 0, 1, 1, 1, 0,
1, 1, 1, 0, 1, 1, 0,
0, 1, 1, 1, 0, 1, 1,
0, 1, 1, 1, 1, 0, 1,
0, 0, 0, 0, 1, 1, 0;
isOn_relation.resize(7,7);
isOn_relation.setZero();
isOn_relation(1,3) = 1;
isOn_relation(2,3) = 1;
coplanar_relation.resize(7,7);
coplanar_relation.setZero();
vector<Eigen::MatrixXi> g4_relations;
g4_relations.push_back( isOn_relation );
g4_relations.push_back( coplanar_relation );
Eigen::VectorXi g4_groundTruth(7);
g4_groundTruth << 1, 6, 9, 2, 5, 4, 0;
fillGraph( graph4, g4_nf, simpleNodeType1Ptr,
g4_adj, simpleEdgeType1Ptr, g4_relations,
g4_groundTruth, groundTruth4 );
graphs.push_back( graph4 );
groundTruths.push_back( groundTruth4 );
///*--------------------------------- Graph 5 ----------------------------------*/
cout << "Graph 5" << endl;
// Desktop 6
CGraph graph5;
std::map<size_t,size_t> groundTruth5;
Eigen::MatrixXd g5_nf(8,5);
g5_nf << 1, 1.10316, 0.363647, 1.61694, 1,
1, 1.00384, 0.23474, 2.7587, 1,
1, 0.988903, 0.118134, 1, 1,
0, 0.789307, 1.62451, 2.07762, 1,
1, 0.672498, 1.8316, 1.78692, 1,
1, 0.976089, 0.244898, 1, 1,
0, 0.502879, 0.186037, 2.29688, 1,
0, 0, 2.50486, 1.43262, 1;
Eigen::MatrixXi g5_adj(8,8);
g5_adj << 0, 1, 1, 1, 1, 1, 0, 0,
1, 0, 1, 1, 1, 0, 0, 0,
1, 1, 0, 1, 0, 0, 0, 0,
1, 1, 1, 0, 1, 1, 1, 0,
1, 1, 0, 1, 0, 1, 1, 1,
1, 0, 0, 1, 1, 0, 1, 0,
0, 0, 0, 1, 1, 1, 0, 0,
0, 0, 0, 0, 1, 0, 0, 0;
isOn_relation.resize(8,8);
isOn_relation.setZero();
isOn_relation(0,3) = 1;
isOn_relation(2,3) = 1;
coplanar_relation.resize(8,8);
coplanar_relation.setZero();
vector<Eigen::MatrixXi> g5_relations;
g5_relations.push_back( isOn_relation );
g5_relations.push_back( coplanar_relation );
Eigen::VectorXi g5_groundTruth(8);
g5_groundTruth << 6, 1, 8, 2, 1, 1, 5, 0;
fillGraph( graph5, g5_nf, simpleNodeType1Ptr,
g5_adj, simpleEdgeType1Ptr, g5_relations,
g5_groundTruth, groundTruth5 );
graphs.push_back( graph5 );
groundTruths.push_back( groundTruth5 );
///*--------------------------------- Graph 6 ----------------------------------*/
cout << "Graph 6" << endl;
// Desktop 7
CGraph graph6;
std::map<size_t,size_t> groundTruth6;
Eigen::MatrixXd g6_nf(7,5);
g6_nf << 1, 0.975437, 1.21382, 2.28555, 1,
1, 0.951433, 0.270196, 2.29035, 1,
0, 0.766445, 1.00624, 1.6777, 1,
0, 0, 1.13219, 1.64994, 1,
1, 0.985712, 0.208164, 1.67471, 1,
0, 0.499913, 0.208579, 1.95149, 1,
1, 0.970966, 0.16066, 1, 1;
Eigen::MatrixXi g6_adj(7,7);
g6_adj << 0, 1, 1, 1, 0, 0, 1,
1, 0, 1, 0, 1, 1, 1,
1, 1, 0, 0, 1, 1, 1,
1, 0, 0, 0, 0, 1, 0,
0, 1, 1, 0, 0, 1, 0,
0, 1, 1, 1, 1, 0, 0,
1, 1, 1, 0, 0, 0, 0;
isOn_relation.resize(7,7);
isOn_relation.setZero();
isOn_relation(2,6) = 1;
coplanar_relation.resize(7,7);
coplanar_relation.setZero();
vector<Eigen::MatrixXi> g6_relations;
g6_relations.push_back( isOn_relation );
g6_relations.push_back( coplanar_relation );
Eigen::VectorXi g6_groundTruth(7);
g6_groundTruth << 1, 1, 2, 0, 4, 5, 6;
fillGraph( graph6, g6_nf, simpleNodeType1Ptr,
g6_adj, simpleEdgeType1Ptr, g6_relations,
g6_groundTruth, groundTruth6 );
graphs.push_back( graph6 );
groundTruths.push_back( groundTruth6 );
///*--------------------------------- Graph 7 ----------------------------------*/
cout << "Graph 7" << endl;
// Desktop 8
CGraph graph7;
std::map<size_t,size_t> groundTruth7;
Eigen::MatrixXd g7_nf(7,5);
g7_nf << 1, 1.08949, 0.44, 1.51331, 1,
1, 1.07257, 0.144868, 1, 1,
0, 0.802159, 1.19101, 1.78961, 1,
1, 0.728912, 0.128203, 2.15983, 1,
0, 0, 3.7626, 2.86856, 1,
0, 0.0142013, 0.175163, 2.24823, 1,
0, 0.540925, 0.205526, 1, 1;
Eigen::MatrixXi g7_adj(7,7);
g7_adj << 0, 1, 1, 1, 0, 0, 1,
1, 0, 1, 0, 0, 0, 0,
1, 1, 0, 1, 0, 0, 1,
1, 0, 1, 0, 0, 0, 1,
0, 0, 0, 0, 0, 1, 1,
0, 0, 0, 0, 1, 0, 1,
1, 0, 1, 1, 1, 1, 0;
isOn_relation.resize(7,7);
isOn_relation.setZero();
isOn_relation(0,2) = 1;
isOn_relation(1,2) = 1;
coplanar_relation.resize(7,7);
coplanar_relation.setZero();
vector<Eigen::MatrixXi> g7_relations;
g7_relations.push_back( isOn_relation );
g7_relations.push_back( coplanar_relation );
Eigen::VectorXi g7_groundTruth(7);
g7_groundTruth << 6, 6, 2, 4, 0, 0, 5;
fillGraph( graph7, g7_nf, simpleNodeType1Ptr,
g7_adj, simpleEdgeType1Ptr, g7_relations,
g7_groundTruth, groundTruth7 );
graphs.push_back( graph7 );
groundTruths.push_back( groundTruth7 );
///*--------------------------------- Graph 8 ----------------------------------*/
cout << "Graph 8" << endl;
// Desktop 9
CGraph graph8;
std::map<size_t,size_t> groundTruth8;
Eigen::MatrixXd g8_nf(5,5);
g8_nf << 1, 1.0398, 1.45393, 1.85835, 1,
0, 0.687862, 1.05335, 1.44245, 1,
0, 0.476001, 0.17945, 1, 1,
1, 0.150004, 0.163443, 1.72806, 1,
0, 0, 2.11644, 3.64183, 1;
Eigen::MatrixXi g8_adj(5,5);
g8_adj << 0, 1, 0, 0, 0,
1, 0, 1, 1, 0,
0, 1, 0, 1, 1,
0, 1, 1, 0, 1,
0, 0, 1, 1, 0;
isOn_relation.resize(5,5);
isOn_relation.setZero();
coplanar_relation.resize(5,5);
coplanar_relation.setZero();
vector<Eigen::MatrixXi> g8_relations;
g8_relations.push_back( isOn_relation );
g8_relations.push_back( coplanar_relation );
Eigen::VectorXi g8_groundTruth(5);
g8_groundTruth << 1, 2, 5, 1, 0;
fillGraph( graph8, g8_nf, simpleNodeType1Ptr,
g8_adj, simpleEdgeType1Ptr, g8_relations,
g8_groundTruth, groundTruth8 );
graphs.push_back( graph8 );
groundTruths.push_back( groundTruth8 );
///*--------------------------------- Graph 9 ----------------------------------*/
cout << "Graph 9" << endl;
// Desktop 11
CGraph graph9;
std::map<size_t,size_t> groundTruth9;
Eigen::MatrixXd g9_nf(6,5);
g9_nf << 1, 1.08032, 0.175311, 1.59692, 1,
1, 1.04474, 0.153065, 1.75199, 1,
0, 0.802788, 1.36259, 2.70243, 1,
1, 0.727563, 0.115821, 1.56757, 1,
0, 0, 2.24629, 1.73137, 1,
0, 0.516504, 0.165597, 1, 1;
Eigen::MatrixXi g9_adj(6,6);
g9_adj << 0, 1, 1, 1, 0, 0,
1, 0, 1, 0, 0, 0,
1, 1, 0, 1, 0, 1,
1, 0, 1, 0, 1, 1,
0, 0, 0, 1, 0, 1,
0, 0, 1, 1, 1, 0;
isOn_relation.resize(6,6);
isOn_relation.setZero();
isOn_relation(0,2) = 1;
coplanar_relation.resize(6,6);
coplanar_relation.setZero();
vector<Eigen::MatrixXi> g9_relations;
g9_relations.push_back( isOn_relation );
g9_relations.push_back( coplanar_relation );
Eigen::VectorXi g9_groundTruth(6);
g9_groundTruth << 6, 1, 2, 4, 0, 5;
fillGraph( graph9, g9_nf, simpleNodeType1Ptr,
g9_adj, simpleEdgeType1Ptr, g9_relations,
g9_groundTruth, groundTruth9 );
graphs.push_back( graph9 );
groundTruths.push_back( groundTruth9 );
///*--------------------------------- Graph 10 ----------------------------------*/
cout << "Graph 10" << endl;
// Desktop 13
CGraph graph10;
std::map<size_t,size_t> groundTruth10;
Eigen::MatrixXd g10_nf(7,5);
g10_nf << 1, 1.13981, 0.613381, 1.46317, 1,
1, 1.01258, 0.167784, 1.47751, 1,
0, 0.772106, 1.55737, 3.7527, 1,
1, 1.11485, 0.201305, 2.29086, 1,
0, 0.487854, 0.161754, 1.53537, 1,
0, 0, 0.70792, 2.77777, 1,
0, 0.0221937, 0.310503, 1.72698, 1;
Eigen::MatrixXi g10_adj(7,7);
g10_adj << 0, 1, 1, 1, 0, 0, 0,
1, 0, 1, 1, 0, 0, 0,
1, 1, 0, 1, 1, 0, 0,
1, 1, 1, 0, 1, 0, 0,
0, 0, 1, 1, 0, 1, 1,
0, 0, 0, 0, 1, 0, 1,
0, 0, 0, 0, 1, 1, 0;
isOn_relation.resize(7,7);
isOn_relation.setZero();
isOn_relation(1,2) = 1;
isOn_relation(2,3) = 1;
coplanar_relation.resize(7,7);
coplanar_relation.setZero();
vector<Eigen::MatrixXi> g10_relations;
g10_relations.push_back( isOn_relation );
g10_relations.push_back( coplanar_relation );
Eigen::VectorXi g10_groundTruth(7);
g10_groundTruth << 1, 6, 2, 8, 5, 0, 0;
fillGraph( graph10, g10_nf, simpleNodeType1Ptr,
g10_adj, simpleEdgeType1Ptr, g10_relations,
g10_groundTruth, groundTruth10 );
graphs.push_back( graph10 );
groundTruths.push_back( groundTruth10 );
///*--------------------------------- Graph 11 ----------------------------------*/
cout << "Graph 11" << endl;
// Desktop 14
CGraph graph11;
std::map<size_t,size_t> groundTruth11;
Eigen::MatrixXd g11_nf(3,5);
g11_nf << 1, 1.39389, 1.07869, 3.04842, 1,
0, 0.724976, 0.627546, 3.18692, 1,
1, 1.02144, 0.168219, 1.53329, 1;
Eigen::MatrixXi g11_adj(3,3);
g11_adj << 0, 1, 1,
1, 0, 1,
1, 1, 0;
isOn_relation.resize(3,3);
isOn_relation.setZero();
isOn_relation(1,2) = 1;
coplanar_relation.resize(3,3);
coplanar_relation.setZero();
vector<Eigen::MatrixXi> g11_relations;
g11_relations.push_back( isOn_relation );
g11_relations.push_back( coplanar_relation );
Eigen::VectorXi g11_groundTruth(3);
g11_groundTruth << 1, 2, 6;
fillGraph( graph11, g11_nf, simpleNodeType1Ptr,
g11_adj, simpleEdgeType1Ptr, g11_relations,
g11_groundTruth, groundTruth11 );
graphs.push_back( graph11 );
groundTruths.push_back( groundTruth11 );
///*--------------------------------- Graph 12 ----------------------------------*/
cout << "Graph 12" << endl;
// Desktop 15
CGraph graph12;
std::map<size_t,size_t> groundTruth12;
Eigen::MatrixXd g12_nf(9,5);
g12_nf << 1, 0.890949, 3.09357, 2.35317, 1,
0, 0.337975, 0.284948, 1.73898, 1,
1, 0.722686, 0.397614, 1.45951, 1,
0, 0.619682, 1.56466, 2.28344, 1,
1, 0.404655, 0.167949, 1, 1,
1, 0.726681, 0.198501, 1.52176, 1,
1, 0.876247, 0.182265, 1.92814, 1,
1, 0.639947, 0.133658, 2.02595, 1,
0, 0, 0.36366, 1, 1;
Eigen::MatrixXi g12_adj(9,9);
g12_adj << 0, 1, 1, 0, 1, 1, 0, 0, 0,
1, 0, 1, 1, 1, 0, 0, 0, 0,
1, 1, 0, 1, 1, 0, 1, 0, 0,
0, 1, 1, 0, 1, 1, 1, 1, 0,
1, 1, 1, 1, 0, 0, 1, 0, 0,
1, 0, 0, 1, 0, 0, 0, 0, 0,
0, 0, 1, 1, 1, 0, 0, 0, 0,
0, 0, 0, 1, 0, 0, 0, 0, 1,
0, 0, 0, 0, 0, 0, 0, 1, 0;
isOn_relation.resize(9,9);
isOn_relation.setZero();
isOn_relation(3,6) = 1;
coplanar_relation.resize(9,9);
coplanar_relation.setZero();
vector<Eigen::MatrixXi> g12_relations;
g12_relations.push_back( isOn_relation );
g12_relations.push_back( coplanar_relation );
Eigen::VectorXi g12_groundTruth(9);
g12_groundTruth << 1, 5, 4, 2, 3, 1, 6, 4, 0;
fillGraph( graph12, g12_nf, simpleNodeType1Ptr,
g12_adj, simpleEdgeType1Ptr, g12_relations,
g12_groundTruth, groundTruth12 );
graphs.push_back( graph12 );
groundTruths.push_back( groundTruth12 );
///*--------------------------------- Graph 13 ----------------------------------*/
// Eigen::MatrixXi g13_adj(11,11);
//, 0, 1, 1, 1, 1, 1, 0, 1, 1, 0
//, 1, 0, 1, 0, 0, 0, 0, 1, 1, 0
//, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0
//, 1, 0, 0, 0, 1, 1, 1, 1, 0, 1
//, 1, 0, 0, 1, 0, 1, 1, 1, 1, 1
//, 1, 0, 0, 1, 1, 0, 1, 0, 0, 0
//, 0, 0, 0, 1, 1, 1, 0, 1, 0, 1
//, 1, 1, 0, 1, 1, 0, 1, 0, 1, 0
//, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0
//, 0, 0, 0, 1, 1, 0, 1, 0, 0, 0
///*--------------------------------- Graph 14 ----------------------------------*/
// Eigen::MatrixXi g14_adj(11,11);
//, 0, 1, 0, 1, 1, 1
//, 1, 0, 1, 1, 0, 0
//, 0, 1, 0, 1, 1, 0
//, 1, 1, 1, 0, 1, 0
//, 1, 0, 1, 1, 0, 0
//, 1, 0, 0, 0, 0, 0
///*--------------------------------- Graph 15 ----------------------------------*/
// Eigen::MatrixXi g15_adj(11,11);
//, 0, 1, 1, 1, 0, 0, 0, 1
//, 1, 0, 1, 1, 0, 0, 0, 0
//, 1, 1, 0, 1, 1, 1, 0, 1
//, 1, 1, 1, 0, 1, 1, 0, 1
//, 0, 0, 1, 1, 0, 1, 1, 0
//, 0, 0, 1, 1, 1, 0, 1, 0
//, 0, 0, 0, 0, 1, 1, 0, 1
//, 1, 0, 1, 1, 0, 0, 1, 0
///*--------------------------------- Graph 16 ----------------------------------*/
// Eigen::MatrixXi g16_adj(11,11);
//, 0, 1, 1, 0, 0, 0
//, 1, 0, 1, 1, 1, 0
//, 1, 1, 0, 1, 1, 0
//, 0, 1, 1, 0, 1, 1
//, 0, 1, 1, 1, 0, 1
//, 0, 0, 0, 1, 1, 0
///*--------------------------------- Graph 17 ----------------------------------*/
// Eigen::MatrixXi g17_adj(11,11);
//, 0, 1, 1, 1, 1, 0, 0
//, 1, 0, 1, 1, 0, 0, 0
//, 1, 1, 0, 1, 1, 1, 0
//, 1, 1, 1, 0, 1, 1, 1
//, 1, 0, 1, 1, 0, 1, 0
//, 0, 0, 1, 1, 1, 0, 0
//, 0, 0, 0, 1, 0, 0, 0
///*--------------------------------- Graph 18 ----------------------------------*/
// Eigen::MatrixXi g18_adj(11,11);
//, 0, 1, 1, 1, 0, 0, 1, 1
//, 1, 0, 1, 0, 1, 1, 1, 0
//, 1, 1, 0, 0, 1, 1, 1, 1
//, 1, 0, 0, 0, 0, 1, 0, 0
//, 0, 1, 1, 0, 0, 1, 0, 0
//, 0, 1, 1, 1, 1, 0, 0, 0
//, 1, 1, 1, 0, 0, 0, 0, 0
//, 1, 0, 1, 0, 0, 0, 0, 0
///*--------------------------------- Graph 19 ----------------------------------*/
// Eigen::MatrixXi g19_adj(11,11);
//, 0, 1, 0, 0, 0, 0
//, 1, 0, 1, 0, 0, 1
//, 0, 1, 0, 0, 0, 1
//, 0, 0, 0, 0, 1, 1
//, 0, 0, 0, 1, 0, 1
//, 0, 1, 1, 1, 1, 0
///*--------------------------------- Graph 20 ----------------------------------*/
// Eigen::MatrixXi g20_adj(11,11);
//, 0, 1, 0, 0, 0
//, 1, 0, 1, 1, 0
//, 0, 1, 0, 1, 1
//, 0, 1, 1, 0, 1
//, 0, 0, 1, 1, 0
///*--------------------------------- Graph 21 ----------------------------------*/
// Eigen::MatrixXi g21_adj(11,11);
//, 0, 1, 1, 1, 0, 0, 0
//, 1, 0, 1, 0, 0, 0, 0
//, 1, 1, 0, 1, 0, 1, 1
//, 1, 0, 1, 0, 1, 1, 0
//, 0, 0, 0, 1, 0, 1, 0
//, 0, 0, 1, 1, 1, 0, 0
//, 0, 0, 1, 0, 0, 0, 0
///*--------------------------------- Graph 22 ----------------------------------*/
// Eigen::MatrixXi g22_adj(11,11);
//, 0, 1, 1, 0, 0
//, 1, 0, 1, 0, 0
//, 1, 1, 0, 1, 0
//, 0, 0, 1, 0, 1
//, 0, 0, 0, 1, 1
///*--------------------------------- Graph 23 ----------------------------------*/
// Eigen::MatrixXi g23_adj(11,11);
//, 0, 1, 1
//, 1, 0, 1
//, 1, 1, 0
///*--------------------------------- Graph 24 ----------------------------------*/
// Eigen::MatrixXi g24_adj(11,11);
// g24_adj << 0, 1, 1, 0, 1, 1, 0, 0, 0,
// 1, 0, 1, 1, 1, 0, 0, 0, 0,
// 1, 1, 0, 1, 1, 0, 1, 0, 0,
// 0, 1, 1, 0, 1, 1, 1, 1, 0,
// 1, 1, 1, 1, 0, 0, 1, 0, 0,
// 1, 0, 0, 1, 0, 0, 0, 0, 0,
// 0, 0, 1, 1, 1, 0, 0, 0, 0,
// 0, 0, 0, 1, 0, 0, 0, 0, 1,
// 0, 0, 0, 0, 0, 0, 0, 1, 0;
///*--------------------------------- Graph 25 ----------------------------------*/
// Eigen::MatrixXi g25_adj(11,11);
// g25_adj << 0, 1, 1, 0, 0,
// 1, 0, 1, 1, 1,
// 1, 1, 0, 0, 0,
// 0, 1, 0, 0, 1,
// 0, 1, 0, 1, 0;
static boost::mt19937 rng;
rng.seed(std::time(0));
size_t N_graphsForTraining = 9;
size_t absoluteSuccess_Greedy = 0;
size_t absoluteSuccess_ICM = 0;
size_t absoluteSuccess_AlphaExpansion = 0;
size_t absoluteSuccess_AlphaBetaSwap = 0;
size_t absoluteSuccess_LBP = 0;
size_t absoluteSuccess_TRP = 0;
size_t absoluteSuccess_RBP = 0;
size_t absoluteNumberOfNodes = 0;
size_t N_repetitions = 100;
for ( size_t rep = 0; rep < N_repetitions; rep++ )
{
vector<size_t> graphs_to_train;
size_t N_graphsAdded = 0;
while ( N_graphsAdded < N_graphsForTraining )
{
boost::uniform_int<> int_generator(0,11);
int rand = int_generator(rng);
if ( std::find(graphs_to_train.begin(),graphs_to_train.end(),rand)
== graphs_to_train.end() )
{
graphs_to_train.push_back(rand);
N_graphsAdded++;
}
}
vector<size_t> graphs_to_test;
for ( size_t i = 0; i < graphs.size(); i++ )
if ( std::find(graphs_to_train.begin(),graphs_to_train.end(), i)
== graphs_to_train.end() )
graphs_to_test.push_back(i);
cout << "Graphs to test: ";
for ( size_t i = 0; i < graphs_to_test.size(); i++)
cout << graphs_to_test[i] << " ";
cout << endl;
// graphs_to_train.push_back(0);
// graphs_to_train.push_back(1);
// graphs_to_train.push_back(2);
// graphs_to_train.push_back(3);
// graphs_to_train.push_back(4);
// graphs_to_train.push_back(5);
// graphs_to_train.push_back(6);
// graphs_to_train.push_back(7);
// graphs_to_train.push_back(8);
// graphs_to_train.push_back(9);
// graphs_to_train.push_back(10);
// graphs_to_test.push_back(11);
for ( size_t i = 0; i < graphs_to_train.size(); i++ )
{
trainingDataset.addGraph( graphs[graphs_to_train[i]] );
trainingDataset.addGraphGroundTruth( groundTruths[graphs_to_train[i]] );
}
/*------------------------------------------------------------------------------
*
* TRAINING!
*
*----------------------------------------------------------------------------*/
UPGMpp::TTrainingOptions to;
to.l2Regularization = true;
to.nodeLambda = 10;
to.edgeLambda = 15;
to.showTrainedWeights = false;
to.showTrainingProgress = false;
cout << "------------------------------------------------------" << endl;
cout << " TRAINING" << endl;
cout << "------------------------------------------------------" << endl;
trainingDataset.setTrainingOptions( to );
int res = trainingDataset.train();
if (res!=0) continue;
/*------------------------------------------------------------------------------
*
* PREPARE A GRAPH FOR PERFORMING DECODING AND INFERENCE
*
*----------------------------------------------------------------------------*/
cout << "------------------------------------------------------" << endl;
cout << " TESTING" << endl;
cout << "------------------------------------------------------" << endl;
for ( size_t i = 0; i < graphs_to_test.size(); i++ )
{
graphs[graphs_to_test[i]].computePotentials();
/*------------------------------------------------------------------------------
*
* DECODING!
*
*----------------------------------------------------------------------------*/
cout << endl;
cout << "------------------------------------------------------" << endl;
cout << " DECODING" << endl;
cout << "------------------------------------------------------" << endl;
CICMInferenceMAP decodeICM;
CICMGreedyInferenceMAP decodeICMGreedy;
CAlphaExpansionInferenceMAP decodeAlphaExpansion;
CAlphaBetaSwapInferenceMAP decodeAlphaBetaSawp;
CLBPInferenceMAP decodeLBP;
CTRPBPInferenceMAP decodeTRPBP;
CRBPInferenceMAP decodeRBP;
double totalSuccess_Greedy = 0;
double totalSuccess_ICM = 0;
double totalSuccess_AlphaExpansion = 0;
double totalSuccess_AlphaBetaSwap = 0;
double totalSuccess_LBP = 0;
double totalSuccess_TRP = 0;
double totalSuccess_RBP = 0;
TInferenceOptions options;
options.maxIterations = 100;
std::map<size_t,size_t> resultsMap;
std::map<size_t,size_t>::iterator it;
//std::cout << " RESULTS ICM " << std::endl;
//std::cout << "-----------------------------------" << std::endl;
decodeICM.setOptions( options );
decodeICM.infer( graphs[graphs_to_test[i]], resultsMap );
for ( it = resultsMap.begin(); it != resultsMap.end(); it++ )
if ( it->second == groundTruths[graphs_to_test[i]][ it->first ])
totalSuccess_ICM++;
//showResults( resultsMap );
//std::cout << "-----------------------------------" << std::endl;
//std::cout << " RESULTS ICM GREEDY" << std::endl;
//std::cout << "-----------------------------------" << std::endl;
decodeICMGreedy.setOptions( options );
decodeICMGreedy.infer( graphs[graphs_to_test[i]], resultsMap );
for ( it = resultsMap.begin(); it != resultsMap.end(); it++ )
if ( it->second == groundTruths[graphs_to_test[i]][ it->first ])
totalSuccess_Greedy++;
//showResults( resultsMap );
//std::cout << "-----------------------------------" << std::endl;
//std::cout << " RESULTS ALPHA-EXPANSIONS" << std::endl;
//std::cout << "-----------------------------------" << std::endl;
options.maxIterations = 10000;
decodeAlphaExpansion.setOptions( options );
decodeAlphaExpansion.infer( graphs[graphs_to_test[i]], resultsMap );
for ( it = resultsMap.begin(); it != resultsMap.end(); it++ )
if ( it->second == groundTruths[graphs_to_test[i]][ it->first ])
totalSuccess_AlphaExpansion++;
//showResults( resultsMap );
//std::cout << "-----------------------------------" << std::endl;
//std::cout << " RESULTS ALPHA-BETA" << std::endl;
//std::cout << "-----------------------------------" << std::endl;
options.maxIterations = 10000;
decodeAlphaBetaSawp.setOptions( options );
decodeAlphaBetaSawp.infer( graphs[graphs_to_test[i]], resultsMap );
for ( it = resultsMap.begin(); it != resultsMap.end(); it++ )
{
if ( it->second == groundTruths[graphs_to_test[i]][ it->first ])
totalSuccess_AlphaBetaSwap++;
}
//showResults( resultsMap );
//std::cout << "-----------------------------------" << std::endl;
//std::cout << " RESULTS LBP" << std::endl;
//std::cout << "-----------------------------------" << std::endl;
options.maxIterations = 10000;
decodeLBP.setOptions( options );
decodeLBP.infer( graphs[graphs_to_test[i]], resultsMap );
for ( it = resultsMap.begin(); it != resultsMap.end(); it++ )
if ( it->second == groundTruths[graphs_to_test[i]][ it->first ])
totalSuccess_LBP++;
//showResults( resultsMap );
//std::cout << "-----------------------------------" << std::endl;
//std::cout << " RESULTS TRP" << std::endl;
//std::cout << "-----------------------------------" << std::endl;
options.maxIterations = 10000;
decodeTRPBP.setOptions( options );
decodeTRPBP.infer( graphs[graphs_to_test[i]], resultsMap );
for ( it = resultsMap.begin(); it != resultsMap.end(); it++ )
if ( it->second == groundTruths[graphs_to_test[i]][ it->first ])
totalSuccess_TRP++;
//showResults( resultsMap );
//std::cout << "-----------------------------------" << std::endl;
//std::cout << " RESULTS RBP" << std::endl;
//std::cout << "-----------------------------------" << std::endl;
options.maxIterations = 10000;
decodeRBP.setOptions( options );
//decodeRBP.infer( graphs[graphs_to_test[i]], resultsMap );
for ( it = resultsMap.begin(); it != resultsMap.end(); it++ )
if ( it->second == groundTruths[graphs_to_test[i]][ it->first ])
totalSuccess_RBP++;
//showResults( resultsMap );
//std::cout << "-----------------------------------" << std::endl;
double totalNumberOfNodes = graphs[graphs_to_test[i]].getNodes().size();
//cout << "Total MaxNodePot success: " << 100*(totalSuccess_MaxNodePot / totalNumberOfNodes) << "%" << endl;
cout << "Total Greedy success: " << 100*(totalSuccess_Greedy / totalNumberOfNodes) << "%" << endl;
cout << "Total ICM success: " << 100*(totalSuccess_ICM / totalNumberOfNodes) << "%" << endl;
//cout << "Total Exact success: " << 100*(totalSuccess_Exact / totalNumberOfNodes) << "%" << endl;
cout << "Total AlphaExpan success: " << 100*(totalSuccess_AlphaExpansion / totalNumberOfNodes) << "%" << endl;
cout << "Total AlphaBetaS success: " << 100*(totalSuccess_AlphaBetaSwap / totalNumberOfNodes) << "%" << endl;
cout << "Total LBP success: " << 100*(totalSuccess_LBP / totalNumberOfNodes) << "%" << endl;
cout << "Total TRP success: " << 100*(totalSuccess_TRP / totalNumberOfNodes) << "%" << endl;
cout << "Total RBP success: " << 100*(totalSuccess_RBP / totalNumberOfNodes) << "%" << endl;
absoluteNumberOfNodes += totalNumberOfNodes;
absoluteSuccess_Greedy += totalSuccess_Greedy;
absoluteSuccess_ICM += totalSuccess_ICM;
absoluteSuccess_AlphaExpansion += totalSuccess_AlphaExpansion;
absoluteSuccess_AlphaBetaSwap += totalSuccess_AlphaBetaSwap;
absoluteSuccess_LBP += totalSuccess_LBP;
absoluteSuccess_TRP += totalSuccess_TRP;
absoluteSuccess_RBP += totalSuccess_RBP;
/*------------------------------------------------------------------------------
*
* INFERENCE!
*
*----------------------------------------------------------------------------*/
/*cout << endl;
cout << "------------------------------------------------------" << endl;
cout << " INFERENCE" << endl;
cout << "------------------------------------------------------" << endl;
std::cout << " LOOPY BELIEF PROPAGATION " << std::endl;
std::cout << "-----------------------------------" << std::endl;
CLBPInference LBPInference;
map<size_t,VectorXd> nodeBeliefs;
map<size_t,MatrixXd> edgeBeliefs;
double logZ;
LBPInference.infer( graphs[graphs_to_test[i]], nodeBeliefs, edgeBeliefs, logZ );
map<size_t,VectorXd>::iterator itNodeBeliefs;
for ( itNodeBeliefs = nodeBeliefs.begin(); itNodeBeliefs != nodeBeliefs.end(); itNodeBeliefs++ )
{
std::cout << "Node id " << itNodeBeliefs->first << " beliefs (marginals): "
<< endl << itNodeBeliefs->second << std::endl << std::endl;
}
map<size_t,MatrixXd>::iterator itEdgeBeliefs;
for ( itEdgeBeliefs = edgeBeliefs.begin(); itEdgeBeliefs != edgeBeliefs.end(); itEdgeBeliefs++ )
{
std::cout << "Edge id " << itEdgeBeliefs->first << " beliefs (marginals): "
<< endl << itEdgeBeliefs->second << std::endl << std::endl;
}
cout << "Log Z : " << logZ << endl;*/
}
}
cout << " FINAL RESULTS: " << endl;
//cout << "Total MaxNodePot success: " << 100*(totalSuccess_MaxNodePot / totalNumberOfNodes) << "%" << endl;
cout << "Total Greedy success: " << 100*((double)absoluteSuccess_Greedy / absoluteNumberOfNodes) << "%" << endl;
cout << "Total ICM success: " << 100*((double)absoluteSuccess_ICM / absoluteNumberOfNodes) << "%" << endl;
//cout << "Total Exact success: " << 100*(totalSuccess_Exact / totalNumberOfNodes) << "%" << endl;
cout << "Total AlphaExpan success: " << 100*((double)absoluteSuccess_AlphaExpansion / absoluteNumberOfNodes) << "%" << endl;
cout << "Total AlphaBetaS success: " << 100*((double)absoluteSuccess_AlphaBetaSwap / absoluteNumberOfNodes) << "%" << endl;
cout << "Total LBP success: " << 100*((double)absoluteSuccess_LBP / absoluteNumberOfNodes) << "%" << endl;
cout << "Total TRP success: " << 100*((double)absoluteSuccess_TRP / absoluteNumberOfNodes) << "%" << endl;
cout << "Total RBP success: " << 100*((double)absoluteSuccess_RBP / absoluteNumberOfNodes) << "%" << endl;
// We are ready to take a beer :)
return 0;
}
int main(int argc, char** argv)
{
int args = 1;
#ifdef USE_OPENCL
if (argc < 10) {
#else
if (argc < 7) {
#endif
std::cout << "Not enough arguments" << std::endl;
system("pause");
return 1;
}
DIM = util::toInt(argv[args++]);
N = util::toInt(argv[args++]);
K = util::toInt(argv[args++]);
ITERATIONS = util::toInt(argv[args++]);
RUNS = util::toInt(argv[args++]);
#ifdef USE_OPENCL
AM_LWS = util::toInt(argv[args++]);
RP_LWS = util::toInt(argv[args++]);
CT_LWS = util::toInt(argv[args++]);
USE_ALL_DEVICES = util::toInt(argv[args++]);
#else
device_count = util::toInt(argv[args++]);
#endif
std::cout << "DIM = " << DIM << std::endl;
std::cout << "N = " << N << std::endl;
std::cout << "K = " << K << std::endl;
std::cout << "ITERATIONS = " << ITERATIONS << std::endl;
std::cout << "RUNS = " << RUNS << std::endl;
#ifdef USE_OPENCL
std::cout << "AM_LWS = " << AM_LWS << std::endl;
std::cout << "RP_LWS = " << RP_LWS << std::endl;
std::cout << "CT_LWS = " << CT_LWS << std::endl;
std::cout << "USE_ALL_DEVICES = " << USE_ALL_DEVICES << std::endl << std::endl;
#else
std::cout << "device_count = " << device_count << std::endl << std::endl;
#endif
#ifdef _WIN32
rng.seed();
srand(GetTickCount());
#else
rng.seed();
srand(getTimeMs());
#endif
u = boost::uniform_real<float>(0.0f, 1000000.0f);
gen = new boost::variate_generator<boost::mt19937&, boost::uniform_real<float> >(rng, u);
#ifdef USE_OPENCL
cl_int clError = CL_SUCCESS;
initCL();
for (int i = 0; i < clDevices.size(); ++i) {
clInputBuf.push_back(cl::Buffer(clContext, CL_MEM_READ_ONLY, N * DIM * sizeof(float), NULL, &clError));
if (clError != CL_SUCCESS) std::cout << "OpenCL Error: Could not create buffer" << std::endl;
clCentroidBuf.push_back(cl::Buffer(clContext, CL_MEM_READ_WRITE, K * DIM * sizeof(float), NULL, &clError));
if (clError != CL_SUCCESS) std::cout << "OpenCL Error: Could not create buffer" << std::endl;
clMappingBuf.push_back(cl::Buffer(clContext, CL_MEM_READ_WRITE, N * sizeof(int), NULL, &clError));
if (clError != CL_SUCCESS) std::cout << "OpenCL Error: Could not create buffer" << std::endl;
clReductionBuf.push_back(cl::Buffer(clContext, CL_MEM_WRITE_ONLY, N * sizeof(float), NULL, &clError));
if (clError != CL_SUCCESS) std::cout << "OpenCL Error: Could not create buffer" << std::endl;
clClusterAssignment[i].setArgs(clInputBuf[i](), clCentroidBuf[i](), clMappingBuf[i]());
clClusterReposition[i].setArgs(clInputBuf[i](), clMappingBuf[i](), clCentroidBuf[i]());
clClusterReposition_k[i].setArgs(clInputBuf[i](), clMappingBuf[i](), clCentroidBuf[i]());
//clClusterReposition_k_c[i].setArgs(clInputBuf[i](), clMappingBuf[i](), clCentroidBuf[i](), clConvergedBuf[i]());
clComputeCost[i].setArgs(clInputBuf[i](), clCentroidBuf[i](), clMappingBuf[i](), clReductionBuf[i]());
}
device_count = clDevices.size();
#endif
util::Clock clock;
clock.reset();
for (int i = 0; i < RUNS; ++i) {
mapping_list.push_back(NULL);
centroids_list.push_back(NULL);
cost_list.push_back(0.0f);
}
float* source = new float[N*DIM];
for (int i = 0; i < N*DIM; ++i)
source[i] = gen_random_float();
input_list.push_back(source);
for (int i = 1; i < device_count; ++i) {
float* copy = new float[N*DIM];
memcpy(copy, source, N*DIM*sizeof(float));
input_list.push_back(copy);
}
if (device_count > 1) {
boost::thread_group threads;
for (int i = 0; i < device_count; ++i) {
threads.create_thread(boost::bind(exec, i, true));
}
threads.join_all();
} else {
exec(0, false);
}
#ifdef USE_OPENCL
reduction_group.join_all();
#endif
int best_result = 0;
float best_cost = std::numeric_limits<float>::max();
for (int i = 0; i < RUNS; ++i) {
if (cost_list[i] < best_cost) {
best_cost = cost_list[i];
best_result = i;
}
}
FILE *out_fdesc = fopen("centroids.out", "wb");
fwrite((void*)centroids_list[best_result], K * DIM * sizeof(float), 1, out_fdesc);
fclose(out_fdesc);
out_fdesc = fopen("mapping.out", "wb");
fwrite((void*)mapping_list[best_result], N * sizeof(int), 1, out_fdesc);
fclose(out_fdesc);
std::cout << "Best result is " << best_result << std::endl;
for (int i = 0; i < device_count; ++i) {
delete[] input_list[i];
}
for (int i = 0; i < RUNS; ++i) {
delete[] mapping_list[i];
delete[] centroids_list[i];
}
float now = clock.get();
std::cout << "Total: " << now << std::endl;
system("pause");
return 0;
}
void modify_seed(unsigned thread)
{
rng.seed(thread);
}
int scheduler::batch_run(int max_thread)
{
// every single line of run stat file
vector<shared_ptr<stringstream> > pvec_stat_line;// dummy file buffer to write,this buffer is to keep the run stat data output FIFO ordered for post-computation analysis convenience
// using pointer to work around the problem that stringstream is non-copyable
{
try
{
// load batch run script file
batch_conf batch_conf;
batch_conf.load_conf(m_batch_conf_path);
vector<shared_ptr<func_base> > vec_ptr_func;// function object base class pointer
vector<shared_ptr<para_base> > vec_ptr_para;// parameter object base class pointer
vector<shared_ptr<alg_base> > vec_ptr_alg;// algorithm object base class pointer
int num_alg;// number of algorithm wait to test
num_alg=batch_conf.tasks.size();
vec_ptr_func.resize(num_alg);
vec_ptr_para.resize(num_alg);
vec_ptr_alg.resize(num_alg);
// pool algo_pool(num_alg<max_thread?num_alg:max_thread);// thread pool of algorithms
pool algo_pool(max_thread); // thread pool of algorithms
int cur_task_count=0;
// batch run of specified algorithm
int alloc_algo_res;
int algo_type;
string conf_path;
bool first_task_run=true;// first time run of batch run
bool seq_task_first_time=true;// first time run of single sequential value task
bool different_algo=true;
int pre_algo_type=0;
int i,j;
vec_task::iterator itr=batch_conf.tasks.begin();
for ( i=0;itr!=batch_conf.tasks.end();++itr,i++ )
{
algo_type=(*itr).algo_code;
conf_path=(*itr).conf_path;
// allocate specific algorithm object dynamically
alloc_algo_res=allocate_alg(algo_type,vec_ptr_alg[i],conf_path);
allocate_para(algo_type,vec_ptr_para[i],conf_path);
int load_res;
// load configuration file data
load_res=vec_ptr_para[i]->read_para(conf_path);
if ( -1==load_res )
{
cout << "read config file ERROR!"
<< "\n";
return -1;
}
different_algo=(pre_algo_type!=algo_type);
if ( different_algo )
seq_task_first_time=true;
int func_type=vec_ptr_para[i]->get_obj_func();// test problem type
int num_dim=vec_ptr_para[i]->get_dim();// dimension of test problem
int stop_type=vec_ptr_para[i]->get_stop_type();// termination criterion value
// allocate Objective Function object dynamically
int alloc_fun_res=allocate_func(func_type,vec_ptr_func[i],num_dim);
bool max_gen_set=has_max_gen(stop_type);
// seed random number generator for EACH task
gen.seed(static_cast<uint32_t>(time(0)));
// generate output file name generic prefix
string str_out_prefix_common;
str_out_prefix_common=gen_out_prefix_common(algo_type,vec_ptr_func[i],vec_ptr_para[i]);
if ( first_task_run ) // print overall batch run stat file title once
{
print_algo_stat_title(run_stat_file,max_gen_set);
print_common_filelist_title(com_filelist);
first_task_run=false;
}
// start sequential parameter valus SPECIFIC processing
bool has_seq_val=vec_ptr_para[i]->has_seq_value();
if ( has_seq_val )
{
// open sequential parameter value stat file BEFORE all combination is scheduled
string para_stat_path=gen_para_stat_file_name(str_out_prefix_common);
// using heap memory and smart pointer to shared the data to sub-threads
locked_para_stat_os lps_os={shared_ptr<ostream>(new ofstream(para_stat_path.c_str())),shared_ptr<mutex>(new mutex)};
shared_ptr<ostream> &ppara_stat_file=lps_os.pos;
if ( seq_task_first_time )
{
print_para_stat_title(*ppara_stat_file,gen_seq_var_name_str(vec_ptr_para[i]),gen_seq_var_num_str(vec_ptr_para[i]),max_gen_set);
print_file_name(para_stat_filelist,func_type,algo_type,para_stat_path);
seq_task_first_time=false;
}
// start sequential paramter value processing
const vector<string>& seq_var_name=vec_ptr_para[i]->get_seq_var_name_vec();
const int seq_var_size=seq_var_name.size();
const vector<int>& seq_dim_size=vec_ptr_para[i]->get_seq_dim_size_vec();
int comb_num=1;
// calc total combination number
for ( j=0;j<seq_var_size;j++ )
comb_num *= seq_dim_size[j];
// generate all parameter value combination index
vector<vector<int> > para_comb_idx;
para_comb_idx.push_back(vector<int>(seq_var_size,0));
int cur_dim;
int last_dim=seq_var_size-1;
// using this method since seq_var_size is NOT constant,
// hence we can't use for loop to iterate
for ( j=1;j<comb_num;j++ )
{
cur_dim=last_dim;
vector<int> cur_comb_idx=para_comb_idx.back();
// find approriate dim to increment index
while ( cur_comb_idx[cur_dim] == seq_dim_size[cur_dim]-1 )
cur_dim--;
cur_comb_idx[cur_dim]++;// 'carry propagation'
if ( cur_dim<last_dim )
{
int j;
for ( j=cur_dim+1;j<seq_var_size;j++ )
cur_comb_idx[j]=0;
}
para_comb_idx.push_back(cur_comb_idx);
}
// start combination schedule
string cur_para;
int val_idx;
vector<int> cur_comb_idx(seq_var_size);
int k;
// allocate EXTRA algorithm objects andd function objects
// Parameters object are SHARED between different parameter value combination(const)
vector<shared_ptr<alg_base> > vec_seq_alg;
vector<shared_ptr<func_base> > vec_seq_func;
vector<shared_ptr<para_base> > vec_seq_para;
vec_seq_alg.resize(comb_num);
vec_seq_func.resize(comb_num);
vec_seq_para.resize(comb_num);
vec_seq_alg[0]=vec_ptr_alg[i];
vec_seq_func[0]=vec_ptr_func[i];
vec_seq_para[0]=vec_ptr_para[i];
for ( k=1;k<comb_num;k++ )
{
allocate_alg(algo_type,vec_seq_alg[k],conf_path);
allocate_func(func_type,vec_seq_func[k],num_dim);
allocate_para(algo_type,vec_seq_para[k],conf_path);
(*vec_seq_para[k])=(*vec_ptr_para[i]);// copy parameter values
}
for ( k=0;k<comb_num;k++ )
{
// alter the sequential parameter value for grouped run
cur_comb_idx=para_comb_idx[k];
int j;
for ( j=0;j<seq_var_size;j++ )
{
cur_para=seq_var_name[j];
val_idx=cur_comb_idx[j];
vec_seq_para[k]->set_cur_seq_val_idx(cur_para,val_idx);
}
// all combination values are set
// set task description text for output
string str_out_prefix_para;
str_out_prefix_para=gen_out_prefix_para(vec_seq_para[k],"_");
vec_seq_alg[k]->set_para_val_desc(gen_seq_val_str(vec_seq_para[k]));
string str_para_stat=gen_out_prefix_para(vec_seq_para[k],",");
// output common stat file path list for post-computation plotting and analysis
common_path com_out_path;// common stat file output
gen_common_file_path(com_out_path,str_out_prefix_common,str_out_prefix_para);
vec_seq_alg[k]->set_com_file_path(com_out_path);
print_common_filelist(com_filelist,func_type,algo_type,com_out_path);
print_para_stat_All_Best_Val_Path(para_stat_all_bst_val_filelist,func_type,str_para_stat,com_out_path.all_bst_val_path);// sequential parameters tuning SPECIFIC
// set algorithm specific stat file name
algo_spec_output(func_type,algo_type,vec_seq_alg[k],str_out_prefix_common,str_out_prefix_para);
// run the algo according to single parameter value combination
vec_seq_alg[k]->set_eval_func(vec_seq_func[k]);
vec_seq_alg[k]->load_parameter(vec_seq_para[k]);// Loading algorithm parameters from outer source
vec_seq_alg[k]->initialize();
// progress_timer p_t;// timer of specified algorithm
cur_task_count++;
output_running_prompt(cur_task_count,vec_seq_para[k]);
pvec_stat_line.push_back(shared_ptr<stringstream>(new stringstream));// expand size by 1
shared_ptr<stringstream> pcur_line=pvec_stat_line.back();
shared_ptr<boost::mutex> pmutex=lps_os.pmut;
algo_pool.schedule(
boost::bind(thread_fun,vec_seq_alg[k],func_type,algo_type,
max_gen_set,has_seq_val,pcur_line,
ppara_stat_file,pmutex )
);
}// for every single value combination
}
else
{
// NO value combination:no need to schedule
common_path com_out_path;
gen_common_file_path(com_out_path,str_out_prefix_common,"");
vec_ptr_alg[i]->set_com_file_path(com_out_path);
print_common_filelist(com_filelist,func_type,algo_type,com_out_path);
algo_spec_output(func_type,algo_type,vec_ptr_alg[i],str_out_prefix_common,"");
// run the algo according to specified parameter value combination
vec_ptr_alg[i]->set_eval_func(vec_ptr_func[i]);
vec_ptr_alg[i]->load_parameter(vec_ptr_para[i]);// Loading algorithm parameters from outer source
vec_ptr_alg[i]->initialize();
// progress_timer p_t;// timer of specified algorithm
cur_task_count++;
output_running_prompt(cur_task_count,vec_ptr_para[i]);
pvec_stat_line.push_back(shared_ptr<stringstream>(new stringstream));// expand size by 1,use heap memory to shared the data between main thread and sub-thread
shared_ptr<stringstream> pcur_line=pvec_stat_line.back();
algo_pool.schedule(
boost::bind(thread_fun, vec_ptr_alg[i], func_type,
algo_type, max_gen_set, has_seq_val,
pcur_line, shared_ptr<ostream>(),
shared_ptr<boost::mutex>()));
}// if ( has_seq_val )
pre_algo_type=algo_type;
}// for every single algorithm task
}// try
catch(bad_cast &bc)
{
cout<<"bad_cast caught in batch_run:"
<< bc.what() << "\n";
return -1;
}
catch(logic_error &le)
{
cout<<"logic_error caught in batch_run:"
<< le.what() << "\n";
return -1;
}
catch(exception& e)
{
cout<<"exception caught in batch_run:"
<<e.what()<< "\n";
return -1;
}
}
// all task are finished
print_stat_file(pvec_stat_line);// keep run stat file output data ordered according to FIFO
return 0;
}// end function batch_run
void UPGMpp::getSpanningTree( CGraph &graph, std::vector<size_t> &tree)
{
// TODO: The efficiency of this method can be improved
// Reset tree
tree.clear();
static boost::mt19937 rng;
static bool firstExecution = true;
if ( firstExecution )
{
rng.seed(time(0));
firstExecution = false;
}
vector<CNodePtr> &v_nodes = graph.getNodes();
vector<CNodePtr> v_nodesToExplore;
map<size_t,size_t> nodesToExploreMap;
size_t N_nodes = v_nodes.size();
//cout << "Random: ";
for ( size_t i_node = 0; i_node < N_nodes; i_node++ )
{
boost::uniform_real<> real_generator(0,1);
real_generator.reset();
double rand = real_generator(rng);
//cout << rand << " ";
if ( rand > 0.25 )
{
v_nodesToExplore.push_back( v_nodes[i_node] );
nodesToExploreMap[ v_nodes[i_node]->getID() ] = v_nodesToExplore.size()-1;
}
}
//cout << endl;
//SHOW_VECTOR_NODES_ID("Nodes to explore: ", v_nodesToExplore)
size_t N_nodesToExplore = v_nodesToExplore.size();
if ( !N_nodesToExplore )
return;
bool nodeAdded = true;
vector<bool> v_nodesAdded( N_nodesToExplore, false );
//
// Randomly select the root node
//
boost::uniform_int<> generator(0,N_nodesToExplore-1);
int rand = generator(rng);
//cout << "Root:" << rand << endl;
v_nodesAdded[rand] = true;
multimap<size_t, CEdgePtr> &edgesF = graph.getEdgesF();
while ( nodeAdded )
{
nodeAdded = false;
for ( size_t i_node = 0; i_node < N_nodesToExplore; i_node++ )
{
//cout << "Node " << i_node << " ID "<< v_nodesToExplore[i_node]->getID() << " Added? " << v_nodesAdded[i_node] << endl;
// Check that the node has not been added to the tree yet
if ( !v_nodesAdded[i_node] )
{
size_t nodeID = v_nodesToExplore[i_node]->getID();
NEIGHBORS_IT neighbors = edgesF.equal_range(nodeID);
for ( multimap<size_t,CEdgePtr>::iterator itNeigbhor = neighbors.first;
itNeigbhor != neighbors.second;
itNeigbhor++ )
{
CEdgePtr edgePtr = (*itNeigbhor).second;
size_t neighborID;
if ( !edgePtr->getNodePosition( nodeID ) )
neighborID = edgePtr->getSecondNodeID();
else
neighborID = edgePtr->getFirstNodeID();
//cout << "Current node ID: " << nodeID << " neighbor: " << neighborID << endl;
if ( v_nodesAdded[nodesToExploreMap[neighborID]] )
{
v_nodesAdded[i_node] = true;
nodeAdded = true;
}
}
}
}
}
//SHOW_VECTOR("Nodes in tree: ", v_nodesAdded)
for ( size_t i_node = 0; i_node < v_nodesToExplore.size(); i_node++ )
{
if ( v_nodesAdded[i_node] )
tree.push_back( v_nodesToExplore[i_node]->getID() );
}
//SHOW_VECTOR("Tree: ", tree)
}