void lle(GArgReader& args)
{
// Load the file and params
GMatrix* pData = loadData(args.pop_string());
Holder<GMatrix> hData(pData);
unsigned int nSeed = getpid() * (unsigned int)time(NULL);
GRand prng(nSeed);
GNeighborFinder* pNF = instantiateNeighborFinder(pData, &prng, args);
Holder<GNeighborFinder> hNF(pNF);
int targetDims = args.pop_uint();
// Parse Options
while(args.size() > 0)
{
if(args.if_pop("-seed"))
prng.setSeed(args.pop_uint());
else
throw Ex("Invalid option: ", args.peek());
}
// Transform the data
GLLE transform(pNF->neighborCount(), targetDims, &prng);
transform.setNeighborFinder(pNF);
GMatrix* pDataAfter = transform.doit(*pData);
Holder<GMatrix> hDataAfter(pDataAfter);
pDataAfter->print(cout);
}
void isomap(GArgReader& args)
{
// Load the file and params
GMatrix* pData = loadData(args.pop_string());
Holder<GMatrix> hData(pData);
unsigned int nSeed = getpid() * (unsigned int)time(NULL);
GRand prng(nSeed);
GNeighborFinder* pNF = instantiateNeighborFinder(pData, &prng, args);
Holder<GNeighborFinder> hNF(pNF);
int targetDims = args.pop_uint();
// Parse Options
bool tolerant = false;
while(args.size() > 0)
{
if(args.if_pop("-seed"))
prng.setSeed(args.pop_uint());
else if(args.if_pop("-tolerant"))
tolerant = true;
else
throw Ex("Invalid option: ", args.peek());
}
// Transform the data
GIsomap transform(pNF->neighborCount(), targetDims, &prng);
transform.setNeighborFinder(pNF);
if(tolerant)
transform.dropDisconnectedPoints();
GMatrix* pDataAfter = transform.reduce(*pData);
Holder<GMatrix> hDataAfter(pDataAfter);
pDataAfter->print(cout);
}
void breadthFirstUnfolding(GArgReader& args)
{
// Load the file and params
GMatrix* pData = loadData(args.pop_string());
Holder<GMatrix> hData(pData);
size_t nSeed = getpid() * (unsigned int)time(NULL);
GRand prng(nSeed);
GNeighborFinder* pNF = instantiateNeighborFinder(pData, &prng, args);
Holder<GNeighborFinder> hNF(pNF);
int targetDims = args.pop_uint();
// Parse Options
size_t reps = 1;
Holder<GMatrix> hControlData(NULL);
while(args.size() > 0)
{
if(args.if_pop("-seed"))
nSeed = args.pop_uint();
else if(args.if_pop("-reps"))
reps = args.pop_uint();
else
throw Ex("Invalid option: ", args.peek());
}
// Transform the data
GBreadthFirstUnfolding transform(reps, pNF->neighborCount(), targetDims);
transform.rand().setSeed(nSeed);
transform.setNeighborFinder(pNF);
GMatrix* pDataAfter = transform.reduce(*pData);
Holder<GMatrix> hDataAfter(pDataAfter);
pDataAfter->print(cout);
}
void ManifoldSculpting(GArgReader& args)
{
// Load the file and params
GMatrix* pData = loadData(args.pop_string());
Holder<GMatrix> hData(pData);
unsigned int nSeed = getpid() * (unsigned int)time(NULL);
GRand prng(nSeed);
GNeighborFinder* pNF = instantiateNeighborFinder(pData, &prng, args);
Holder<GNeighborFinder> hNF(pNF);
size_t targetDims = args.pop_uint();
// Parse Options
const char* szPreprocessedData = NULL;
double scaleRate = 0.999;
while(args.size() > 0)
{
if(args.if_pop("-seed"))
prng.setSeed(args.pop_uint());
else if(args.if_pop("-continue"))
szPreprocessedData = args.pop_string();
else if(args.if_pop("-scalerate"))
scaleRate = args.pop_double();
else
throw Ex("Invalid option: ", args.peek());
}
// Load the hint data
GMatrix* pDataHint = NULL;
Holder<GMatrix> hDataHint(NULL);
if(szPreprocessedData)
{
pDataHint = loadData(szPreprocessedData);
hDataHint.reset(pDataHint);
if(pDataHint->relation()->size() != targetDims)
throw Ex("Wrong number of dims in the hint data");
if(pDataHint->rows() != pData->rows())
throw Ex("Wrong number of patterns in the hint data");
}
// Transform the data
GManifoldSculpting transform(pNF->neighborCount(), targetDims, &prng);
transform.setSquishingRate(scaleRate);
if(pDataHint)
transform.setPreprocessedData(hDataHint.release());
transform.setNeighborFinder(pNF);
GMatrix* pDataAfter = transform.doit(*pData);
Holder<GMatrix> hDataAfter(pDataAfter);
pDataAfter->print(cout);
}
void blendEmbeddings(GArgReader& args)
{
// Load the files and params
GMatrix* pDataOrig = loadData(args.pop_string());
Holder<GMatrix> hDataOrig(pDataOrig);
unsigned int seed = getpid() * (unsigned int)time(NULL);
GRand prng(seed);
GNeighborFinder* pNF = instantiateNeighborFinder(pDataOrig, &prng, args);
Holder<GNeighborFinder> hNF(pNF);
GMatrix* pDataA = loadData(args.pop_string());
Holder<GMatrix> hDataA(pDataA);
GMatrix* pDataB = loadData(args.pop_string());
Holder<GMatrix> hDataB(pDataB);
if(pDataA->rows() != pDataOrig->rows() || pDataB->rows() != pDataOrig->rows())
throw Ex("mismatching number of rows");
if(pDataA->cols() != pDataB->cols())
throw Ex("mismatching number of cols");
// Parse Options
while(args.size() > 0)
{
if(args.if_pop("-seed"))
prng.setSeed(args.pop_uint());
else
throw Ex("Invalid option: ", args.peek());
}
// Get a neighbor table
if(!pNF->isCached())
{
GNeighborFinderCacheWrapper* pNF2 = new GNeighborFinderCacheWrapper(hNF.release(), true);
hNF.reset(pNF2);
pNF = pNF2;
}
((GNeighborFinderCacheWrapper*)pNF)->fillCache();
size_t* pNeighborTable = ((GNeighborFinderCacheWrapper*)pNF)->cache();
// Do the blending
size_t startPoint = (size_t)prng.next(pDataA->rows());
double* pRatios = new double[pDataA->rows()];
ArrayHolder<double> hRatios(pRatios);
GVec::setAll(pRatios, 0.5, pDataA->rows());
GMatrix* pDataC = GManifold::blendEmbeddings(pDataA, pRatios, pDataB, pNF->neighborCount(), pNeighborTable, startPoint);
Holder<GMatrix> hDataC(pDataC);
pDataC->print(cout);
}
// virtual
void GNeighborTransducer::transduce(GData* pDataLabeled, GData* pDataUnlabeled, int labelDims)
{
if(labelDims != 1)
ThrowError("Only 1 nominal label is supported");
if(!pDataLabeled->relation()->areNominal(pDataLabeled->relation()->size() - 1, 1))
ThrowError("Only nominal labels are supported");
if(!pDataLabeled->relation()->areContinuous(0, pDataLabeled->relation()->size() - 1))
ThrowError("Only continuous features are supported");
if(pDataLabeled->cols() != pDataUnlabeled->cols())
ThrowError("relations don't match");
// Make a dataset containing all rows
GData dataAll(pDataLabeled->relation());
dataAll.reserve(pDataLabeled->rows() + pDataUnlabeled->rows());
GReleaseDataHolder hDataAll(&dataAll);
for(size_t i = 0; i < pDataUnlabeled->rows(); i++)
dataAll.takeRow(pDataUnlabeled->row(i));
for(size_t i = 0; i < pDataLabeled->rows(); i++)
dataAll.takeRow(pDataLabeled->row(i));
int featureDims = pDataLabeled->cols() - labelDims;
sp_relation pRelInputs = new GUniformRelation(featureDims, 0);
dataAll.setRelation(pRelInputs);
// Find friends
GNeighborFinder* pNF;
if(m_intrinsicDims == 0)
pNF = new GNeighborFinderCacheWrapper(new GKdTree(&dataAll, 0, m_friendCount, NULL, true), true);
else
pNF = new GManifoldNeighborFinder(
&dataAll,
m_friendCount, // littleK
m_friendCount * 4, // bigK
m_intrinsicDims, // intrinsicDims
m_alpha, // alpha
m_beta, // beta
false, // prune?
m_pRand);
Holder<GNeighborFinder> hNF(pNF);
GTEMPBUF(size_t, neighbors, m_friendCount);
int labelValues = pDataLabeled->relation()->valueCount(featureDims);
GTEMPBUF(double, tallys, labelValues);
// Label the unlabeled patterns
GBitTable labeled(pDataUnlabeled->rows());
GData labelList(3); // pattern index, most likely label, confidence
labelList.newRows(pDataUnlabeled->rows());
for(size_t i = 0; i < pDataUnlabeled->rows(); i++)
labelList.row(i)[0] = i;
while(labelList.rows() > 0)
{
// Compute the most likely label and the confidence for each pattern
for(size_t i = 0; i < labelList.rows(); i++)
{
// Find the most common label
double* pRow = labelList.row(i);
size_t index = (size_t)pRow[0];
pNF->neighbors(neighbors, index);
GVec::setAll(tallys, 0.0, labelValues);
for(int j = 0; j < m_friendCount; j++)
{
if(neighbors[j] >= dataAll.rows())
continue;
double* pFriend = dataAll.row(neighbors[j]);
if(neighbors[j] >= pDataUnlabeled->rows())
{
if((int)pFriend[featureDims] >= 0 && (int)pFriend[featureDims] < labelValues)
tallys[(int)pFriend[featureDims]] += 1.0;
}
else if(labeled.bit(neighbors[j]))
{
if((int)pFriend[featureDims] >= 0 && (int)pFriend[featureDims] < labelValues)
tallys[(int)pFriend[featureDims]] += 0.6;
}
}
int label = GVec::indexOfMax(tallys, labelValues, m_pRand);
double conf = tallys[label];
// Penalize for dissenting votes
for(int j = 0; j < m_friendCount; j++)
{
if(neighbors[j] >= dataAll.rows())
continue;
double* pFriend = dataAll.row(neighbors[j]);
if(neighbors[j] >= pDataUnlabeled->rows())
{
if((int)pFriend[featureDims] != label)
conf *= 0.5;
}
else if(labeled.bit(neighbors[j]))
{
if((int)pFriend[featureDims] != label)
conf *= 0.8;
}
}
pRow[1] = label;
pRow[2] = conf;
}
labelList.sort(2);
// Assign the labels to the patterns we are most confident about
size_t maxCount = MAX((size_t)5, pDataLabeled->rows() / 5);
size_t count = 0;
for(size_t i = labelList.rows() - 1; i < labelList.rows(); i--)
{
double* pRow = labelList.row(i);
size_t index = (size_t)pRow[0];
int label = (int)pRow[1];
pDataUnlabeled->row(index)[featureDims] = label;
labeled.set(index);
labelList.deleteRow(i);
if(count >= maxCount)
break;
count++;
}
}
}
