int main(int argc, char** argv)
{
FTAGArgs args;
args.getArgs(argc, argv);
srand48(args._seed);
cout << "SEED=" << args._seed << endl;
ofstream ofile(args._outFile.c_str());
if (!ofile)
{
cerr << "Specifiy output file with ofile=<path>" << endl;
exit(1);
}
ofstream pfile(args._fpFileOut.c_str());
AxtReader ar;
try
{
ar.read(args._inFile);
const vector<AxtPair>& wholeThing = ar.getAlignments();
Pairs allAlignments;
for (size_t i = 0; i < wholeThing.size(); ++i)
{
allAlignments.push_back(wholeThing[i].getAlignment());
}
Pairs alignments = ar.sample(args._numPairs, args._maxLength / 2,
args._maxLength, 0, false);
FTAGParams parAll = estParams(allAlignments, args._symmetric);
FTAGParams parInput = estParams(alignments, args._symmetric);
cout << "From Whole File: " << parAll << endl;
cout << "From input: " << parInput << endl;
ofile << alignments;
if (pfile.is_open())
{
pfile << parInput;
}
}
catch(string message)
{
cerr << message << endl;
return 1;
}
return 0;
}
int main(int argc, char** argv)
{
FTAGArgs args;
args.getArgs(argc, argv);
srand48(args._seed);
cout << "SEED=" << args._seed << endl;
FTAGParams params;
ifstream pfile(args._fpFile.c_str());
if (!pfile)
{
cout << "Specify params with fpfile=<path>" << endl;
}
pfile >> params;
ContextModel em;
em.setDoubleJC(params);
TransitionModel tm;
tm.setSimplified(params);
FTAGGen sGen;
sGen.setEmissionModel(em);
sGen.setTransitionModel(tm);
string a,b;
deque<Trace> trace;
sGen.genAlignment(a, b, trace, true, 10000);
if (drand48() < args._flipProb)
{
reverse(a.begin(), a.end());
reverse(b.begin(), b.end());
}
cout << a << endl << b << endl;
// copy(trace.begin(), trace.end(), ostream_iterator<Trace>(cout, "\n"));
for (deque<Trace>::iterator i = trace.begin(); i != trace.end(); ++i)
cout << *i << endl;
cout << trace.size() << " state transitions used" << endl;
return 0;
}
int main(int argc, char** argv)
{
FTAGArgs args;
args.getArgs(argc, argv);
srand48(args._seed);
cout << "SEED=" << args._seed << endl;
FTAGParams params;
try
{
ifstream pfile(args._fpFile.c_str());
if (!pfile)
{
cout << "Specify params with fpfile=<path>" << endl;
}
pfile >> params;
}
catch(string message)
{
cout << message << endl;
return 1;
}
ContextModel em;
em.setDoubleJC(params);
TransitionModel tm;
tm.setSimplified(params);
AGDOptimizer optimizer;
optimizer.setFixed(params);
optimizer.setModels(em, tm);
FTAGParams offset(params);
offset.setAll(args._offset);
FTAGParams parSeed(params);
parSeed.setSymmetric(args._symmetric);
parSeed.randomizeNonFixed();
optimizer.setSavedInit(parSeed);
optimizer.setRelative(args._relative);
optimizer.setRandomOrder(args._randOrder);
optimizer.setLoopParams(args._numOptTrials, args._maxOptIt,
args._optThreshold, offset);
double mse = optimizer.optimize();
double mymse = params.mse(optimizer.getParams());
TransitionModel tmEst;
tmEst.setSimplified(optimizer.getParams());
cout << "ORIGINAL\n" << tm << "\nESTIMATED\n" << tmEst << endl << endl;
cout << "ORIGINAL\n" << params
<< "\nNEW\n" << optimizer.getParams()
<< "\nDELTA\n" << params - optimizer.getParams()
<< endl;
cout << "MSE=" << mse << endl;
return 0;
}
int main(int argc, char** argv)
{
FTAGArgs args;
args.getArgs(argc, argv);
srand48(args._seed);
cout << "SEED=" << args._seed << endl;
ofstream logFile;
FTAGTrain::EMTrace emTrace;
if (!args._outFile.empty())
{
logFile.open(args._outFile.c_str());
}
ofstream distFile;
vector<size_t> dist, dist_nc;
if (!args._distFile.empty())
{
distFile.open(args._distFile.c_str());
dist = vector<size_t>(args._maxLength, 0);
dist_nc = vector<size_t>(args._maxLength, 0);
}
AxtReader ar;
FTAGParams parSeed;
try
{
ar.read(args._inFile);
ifstream pfile(args._fpFile.c_str());
if (!pfile)
{
cout << "Specify params with fpfile=<path>" << endl;
}
pfile >> parSeed;
}
catch(string message)
{
cout << message << endl;
return 1;
}
if (args._context == false)
{
parSeed.setRMD(0.);
parSeed.setRMI(0.);
parSeed.setPCD(1.);
parSeed.setPCI(1.);
parSeed.setRMDFixed(true);
parSeed.setRMIFixed(true);
parSeed.setPCDFixed(true);
parSeed.setPCIFixed(true);
}
// vector<pair<string, string> > alignments = filter(ar.getAlignments(),
// minLength, maxLength);
const std::vector<AxtPair>& wholeThing = ar.getAlignments();
vector<pair<string, string> > allAlignments;
for (size_t i = 0; i < wholeThing.size(); ++i)
{
allAlignments.push_back(wholeThing[i].getAlignment());
}
vector<pair<string, string> > alignments
= ar.sample(args._numPairs, args._maxLength / 2, args._maxLength, 0,
false);
for (size_t i = 0; i < alignments.size(); ++i)
{
cout << i << "a) " << alignments[i].first << endl
<< i << "b) " << alignments[i].second << endl;
}
FTAGParams parAll = estParams(allAlignments);
FTAGParams parInput = estParams(alignments);
ContextModel emInput;
emInput.setDoubleJC(parInput);
TransitionModel tmInput;
tmInput.setSimplified(parInput);
parSeed.setSymmetric(args._symmetric);
parSeed.setUniGap(args._unigap);
parSeed.randomizeNonFixed();
cout << "From Whole File: " << parAll << endl;
cout << "From input: " << parInput << endl;
cout << "Random Seed: " << parSeed << endl;
FTAGParams offset;
offset.setAll(args._offset);
AGDOptimizer optimizer;
optimizer.setRelative(args._relative);
optimizer.setSavedInit(parSeed);
optimizer.setLoopParams(args._numOptTrials, args._maxOptIt,
args._optThreshold, offset);
EmissionEstimator ee;
TransitionEstimator te;
ee.setBias(args._emBias);
te.setBias(args._tmBias);
FTAGTrain* trainer = new FTAGTrain();
trainer->initialize(optimizer, ee, te);
trainer->setSequences(alignments, args._winSize);
trainer->emLoop(args._maxEMIt, args._emThreshold, args._emConvRepeats,
args._emCRThreshold, &emTrace);
FTAGParams params = trainer->getParams();
TransitionModel tm;
ContextModel em;
tm.setSimplified(params);
em.setDoubleJC(params);
delete trainer;
trainer = NULL;
double pr = 0.;
double prInput = 0.;
double tdiff = 0;
size_t tlength = 0;
for (size_t i = 0; i < alignments.size(); ++i)
{
FTAGModel ftag;
ftag.setSequences(alignments[i].first, alignments[i].second,
args._winSize, args._winSize);
ftag.setTransitionModel(tm);
ftag.setEmissionModel(em);
pr += log(ftag.forward());
string a,b;
deque<Trace> trace;
ftag.viterbi();
ftag.viterbiTrace(a, b, trace);
cout << i << "a) " << a << endl
<< i << "b) " << b << endl;
AxtPair apEstimated(a, b);
AxtPair apOriginal(alignments[i].first, alignments[i].second);
size_t adiff = apOriginal.diff(apEstimated);
ftag.setTransitionModel(tmInput);
ftag.setEmissionModel(emInput);
prInput += log(ftag.forward());
tdiff += adiff;
tlength += alignments[i].first.length();
makeLenDistSimpleSymmetric(trace, dist, dist_nc);
}
FTAGParams parDelta = parInput - params;
parDelta = parDelta.abs();
double parMSE = params.mse(parInput);
cout << "Pr Input= " << prInput << "\nPar_Input= " << parInput << endl
<< "Par_Delta= " << parDelta << "\nPar_MSE= " << parMSE << endl
<< "Num Alignments=" << alignments.size()
<< " Length=[" << args._maxLength/2 << ", " << args._maxLength << "]"
<< " Pr = " << pr << " tdiff = " << tdiff
<< "\nParamEst =\n" << params << endl;
if (logFile)
makeEmGraph(emTrace, prInput, params, tlength, logFile);
if (distFile)
{
distFile << "len freq tot freqNC totNC\n";
for (size_t i = 1; i < dist.size(); ++i)
{
distFile << i << " " << dist[i] << " " << dist[0] << " "
<< dist_nc[i] << " " << dist_nc[0] << "\n";
}
}
return 0;
}
