template<typename SparseMatrixType> void sparse_product()
{
typedef typename SparseMatrixType::Index Index;
Index n = 100;
const Index rows = internal::random<int>(1,n);
const Index cols = internal::random<int>(1,n);
const Index depth = internal::random<int>(1,n);
typedef typename SparseMatrixType::Scalar Scalar;
enum { Flags = SparseMatrixType::Flags };
double density = (std::max)(8./(rows*cols), 0.1);
typedef Matrix<Scalar,Dynamic,Dynamic> DenseMatrix;
typedef Matrix<Scalar,Dynamic,1> DenseVector;
typedef Matrix<Scalar,1,Dynamic> RowDenseVector;
typedef SparseVector<Scalar,0,Index> ColSpVector;
typedef SparseVector<Scalar,RowMajor,Index> RowSpVector;
Scalar s1 = internal::random<Scalar>();
Scalar s2 = internal::random<Scalar>();
// test matrix-matrix product
{
DenseMatrix refMat2 = DenseMatrix::Zero(rows, depth);
DenseMatrix refMat2t = DenseMatrix::Zero(depth, rows);
DenseMatrix refMat3 = DenseMatrix::Zero(depth, cols);
DenseMatrix refMat3t = DenseMatrix::Zero(cols, depth);
DenseMatrix refMat4 = DenseMatrix::Zero(rows, cols);
DenseMatrix refMat4t = DenseMatrix::Zero(cols, rows);
DenseMatrix refMat5 = DenseMatrix::Random(depth, cols);
DenseMatrix refMat6 = DenseMatrix::Random(rows, rows);
DenseMatrix dm4 = DenseMatrix::Zero(rows, rows);
// DenseVector dv1 = DenseVector::Random(rows);
SparseMatrixType m2 (rows, depth);
SparseMatrixType m2t(depth, rows);
SparseMatrixType m3 (depth, cols);
SparseMatrixType m3t(cols, depth);
SparseMatrixType m4 (rows, cols);
SparseMatrixType m4t(cols, rows);
SparseMatrixType m6(rows, rows);
initSparse(density, refMat2, m2);
initSparse(density, refMat2t, m2t);
initSparse(density, refMat3, m3);
initSparse(density, refMat3t, m3t);
initSparse(density, refMat4, m4);
initSparse(density, refMat4t, m4t);
initSparse(density, refMat6, m6);
// int c = internal::random<int>(0,depth-1);
// sparse * sparse
VERIFY_IS_APPROX(m4=m2*m3, refMat4=refMat2*refMat3);
VERIFY_IS_APPROX(m4=m2t.transpose()*m3, refMat4=refMat2t.transpose()*refMat3);
VERIFY_IS_APPROX(m4=m2t.transpose()*m3t.transpose(), refMat4=refMat2t.transpose()*refMat3t.transpose());
VERIFY_IS_APPROX(m4=m2*m3t.transpose(), refMat4=refMat2*refMat3t.transpose());
VERIFY_IS_APPROX(m4 = m2*m3/s1, refMat4 = refMat2*refMat3/s1);
VERIFY_IS_APPROX(m4 = m2*m3*s1, refMat4 = refMat2*refMat3*s1);
VERIFY_IS_APPROX(m4 = s2*m2*m3*s1, refMat4 = s2*refMat2*refMat3*s1);
VERIFY_IS_APPROX(m4=(m2*m3).pruned(0), refMat4=refMat2*refMat3);
VERIFY_IS_APPROX(m4=(m2t.transpose()*m3).pruned(0), refMat4=refMat2t.transpose()*refMat3);
VERIFY_IS_APPROX(m4=(m2t.transpose()*m3t.transpose()).pruned(0), refMat4=refMat2t.transpose()*refMat3t.transpose());
VERIFY_IS_APPROX(m4=(m2*m3t.transpose()).pruned(0), refMat4=refMat2*refMat3t.transpose());
// test aliasing
m4 = m2; refMat4 = refMat2;
VERIFY_IS_APPROX(m4=m4*m3, refMat4=refMat4*refMat3);
// sparse * dense
VERIFY_IS_APPROX(dm4=m2*refMat3, refMat4=refMat2*refMat3);
VERIFY_IS_APPROX(dm4=m2*refMat3t.transpose(), refMat4=refMat2*refMat3t.transpose());
VERIFY_IS_APPROX(dm4=m2t.transpose()*refMat3, refMat4=refMat2t.transpose()*refMat3);
VERIFY_IS_APPROX(dm4=m2t.transpose()*refMat3t.transpose(), refMat4=refMat2t.transpose()*refMat3t.transpose());
VERIFY_IS_APPROX(dm4=m2*(refMat3+refMat3), refMat4=refMat2*(refMat3+refMat3));
VERIFY_IS_APPROX(dm4=m2t.transpose()*(refMat3+refMat5)*0.5, refMat4=refMat2t.transpose()*(refMat3+refMat5)*0.5);
// dense * sparse
VERIFY_IS_APPROX(dm4=refMat2*m3, refMat4=refMat2*refMat3);
VERIFY_IS_APPROX(dm4=refMat2*m3t.transpose(), refMat4=refMat2*refMat3t.transpose());
VERIFY_IS_APPROX(dm4=refMat2t.transpose()*m3, refMat4=refMat2t.transpose()*refMat3);
VERIFY_IS_APPROX(dm4=refMat2t.transpose()*m3t.transpose(), refMat4=refMat2t.transpose()*refMat3t.transpose());
// sparse * dense and dense * sparse outer product
test_outer<SparseMatrixType,DenseMatrix>::run(m2,m4,refMat2,refMat4);
VERIFY_IS_APPROX(m6=m6*m6, refMat6=refMat6*refMat6);
// sparse matrix * sparse vector
ColSpVector cv0(cols), cv1;
DenseVector dcv0(cols), dcv1;
initSparse(2*density,dcv0, cv0);
RowSpVector rv0(depth), rv1;
RowDenseVector drv0(depth), drv1(rv1);
initSparse(2*density,drv0, rv0);
VERIFY_IS_APPROX(cv1=rv0*m3, dcv1=drv0*refMat3);
VERIFY_IS_APPROX(rv1=rv0*m3, drv1=drv0*refMat3);
VERIFY_IS_APPROX(cv1=m3*cv0, dcv1=refMat3*dcv0);
VERIFY_IS_APPROX(cv1=m3t.adjoint()*cv0, dcv1=refMat3t.adjoint()*dcv0);
VERIFY_IS_APPROX(rv1=m3*cv0, drv1=refMat3*dcv0);
}
// test matrix - diagonal product
{
DenseMatrix refM2 = DenseMatrix::Zero(rows, cols);
DenseMatrix refM3 = DenseMatrix::Zero(rows, cols);
DenseMatrix d3 = DenseMatrix::Zero(rows, cols);
DiagonalMatrix<Scalar,Dynamic> d1(DenseVector::Random(cols));
DiagonalMatrix<Scalar,Dynamic> d2(DenseVector::Random(rows));
SparseMatrixType m2(rows, cols);
SparseMatrixType m3(rows, cols);
initSparse<Scalar>(density, refM2, m2);
initSparse<Scalar>(density, refM3, m3);
VERIFY_IS_APPROX(m3=m2*d1, refM3=refM2*d1);
VERIFY_IS_APPROX(m3=m2.transpose()*d2, refM3=refM2.transpose()*d2);
VERIFY_IS_APPROX(m3=d2*m2, refM3=d2*refM2);
VERIFY_IS_APPROX(m3=d1*m2.transpose(), refM3=d1*refM2.transpose());
// also check with a SparseWrapper:
DenseVector v1 = DenseVector::Random(cols);
DenseVector v2 = DenseVector::Random(rows);
VERIFY_IS_APPROX(m3=m2*v1.asDiagonal(), refM3=refM2*v1.asDiagonal());
VERIFY_IS_APPROX(m3=m2.transpose()*v2.asDiagonal(), refM3=refM2.transpose()*v2.asDiagonal());
VERIFY_IS_APPROX(m3=v2.asDiagonal()*m2, refM3=v2.asDiagonal()*refM2);
VERIFY_IS_APPROX(m3=v1.asDiagonal()*m2.transpose(), refM3=v1.asDiagonal()*refM2.transpose());
VERIFY_IS_APPROX(m3=v2.asDiagonal()*m2*v1.asDiagonal(), refM3=v2.asDiagonal()*refM2*v1.asDiagonal());
// evaluate to a dense matrix to check the .row() and .col() iterator functions
VERIFY_IS_APPROX(d3=m2*d1, refM3=refM2*d1);
VERIFY_IS_APPROX(d3=m2.transpose()*d2, refM3=refM2.transpose()*d2);
VERIFY_IS_APPROX(d3=d2*m2, refM3=d2*refM2);
VERIFY_IS_APPROX(d3=d1*m2.transpose(), refM3=d1*refM2.transpose());
}
// test self adjoint products
{
DenseMatrix b = DenseMatrix::Random(rows, rows);
DenseMatrix x = DenseMatrix::Random(rows, rows);
DenseMatrix refX = DenseMatrix::Random(rows, rows);
DenseMatrix refUp = DenseMatrix::Zero(rows, rows);
DenseMatrix refLo = DenseMatrix::Zero(rows, rows);
DenseMatrix refS = DenseMatrix::Zero(rows, rows);
SparseMatrixType mUp(rows, rows);
SparseMatrixType mLo(rows, rows);
SparseMatrixType mS(rows, rows);
do {
initSparse<Scalar>(density, refUp, mUp, ForceRealDiag|/*ForceNonZeroDiag|*/MakeUpperTriangular);
} while (refUp.isZero());
refLo = refUp.adjoint();
mLo = mUp.adjoint();
refS = refUp + refLo;
refS.diagonal() *= 0.5;
mS = mUp + mLo;
// TODO be able to address the diagonal....
for (int k=0; k<mS.outerSize(); ++k)
for (typename SparseMatrixType::InnerIterator it(mS,k); it; ++it)
if (it.index() == k)
it.valueRef() *= 0.5;
VERIFY_IS_APPROX(refS.adjoint(), refS);
VERIFY_IS_APPROX(mS.adjoint(), mS);
VERIFY_IS_APPROX(mS, refS);
VERIFY_IS_APPROX(x=mS*b, refX=refS*b);
VERIFY_IS_APPROX(x=mUp.template selfadjointView<Upper>()*b, refX=refS*b);
VERIFY_IS_APPROX(x=mLo.template selfadjointView<Lower>()*b, refX=refS*b);
VERIFY_IS_APPROX(x=mS.template selfadjointView<Upper|Lower>()*b, refX=refS*b);
// sparse selfadjointView * sparse
SparseMatrixType mSres(rows,rows);
VERIFY_IS_APPROX(mSres = mLo.template selfadjointView<Lower>()*mS,
refX = refLo.template selfadjointView<Lower>()*refS);
// sparse * sparse selfadjointview
VERIFY_IS_APPROX(mSres = mS * mLo.template selfadjointView<Lower>(),
refX = refS * refLo.template selfadjointView<Lower>());
}
}
void testStlDriver()
{
const int numPatches = 200000; // 200000
const int patchWidth = 9;
const int numFeatures = 50;
const double lambda = 0.0005f;
const double epsilon = 1e-2;
Config config;
config.setValue("addBiasTerm", false);
config.setValue("meanStddNormalize", false);
config.setValue("configurePolicyTesting", false);
config.setValue("trainingMeanAndStdd", false);
updateMNISTConfig(config);
if (false)
{
MNISTSamplePatchesUnlabeledDataFunction mnistUnlabeled(numPatches, patchWidth);
SoftICACostFunction sfc(numFeatures, lambda, epsilon);
LIBLBFGSOptimizer lbfgs(200); // 1000
Driver drv1(&config, &mnistUnlabeled, &sfc, &lbfgs);
const Vector_t optThetaRica = drv1.drive();
Matrix_t Wrica(
Eigen::Map<const Matrix_t>(optThetaRica.data(), numFeatures, pow(patchWidth, 2)));
std::ofstream ofs_wrica("../W2.txt");
ofs_wrica << Wrica << std::endl;
}
Matrix_t Wrica;
// debug: read off the values
std::ifstream in("/home/sam/School/online/stanford_dl_ex/W2.txt");
if (in.is_open())
{
std::string str;
int nbRows = 0;
while (std::getline(in, str))
{
if (str.size() == 0)
continue;
std::istringstream iss(str);
std::vector<double> tokens //
{ std::istream_iterator<double> { iss }, std::istream_iterator<double> { } };
Wrica.conservativeResize(nbRows + 1, tokens.size());
for (size_t i = 0; i < tokens.size(); ++i)
Wrica(nbRows, i) = tokens[i];
++nbRows;
}
}
else
{
std::cerr << "file W.txt failed" << std::endl;
exit(EXIT_FAILURE);
}
const int imageDim = 28;
Eigen::Vector2i imageConfig;
imageConfig << imageDim, imageDim;
const int numFilters = numFeatures;
const int poolDim = 5;
const int filterDim = patchWidth;
const int convDim = (imageDim - filterDim + 1);
assert(convDim % poolDim == 0);
const int outputDim = (convDim / poolDim);
StlFilterFunction stlFilterFunction(filterDim, Wrica);
SigmoidFunction sigmoidFunction;
ConvolutionFunction convolutionFunction(&stlFilterFunction, &sigmoidFunction);
MeanPoolFunction meanPoolFunction(numFilters, outputDim);
MNISTSamplePatchesLabeledDataFunction mnistLabeled(&convolutionFunction, &meanPoolFunction,
imageConfig, numFilters, poolDim, outputDim);
SoftmaxCostFunction mnistcf(0.01f);
LIBLBFGSOptimizer lbfgs2(300);
config.setValue("configurePolicyTesting", false);
config.setValue("trainingMeanAndStdd", true);
config.setValue("meanStddNormalize", true);
config.setValue("addBiasTerm", true);
config.setValue("numGrd", true);
config.setValue("training_accuracy", true);
config.setValue("testing_accuracy", true);
//config.setValue("addBiasTerm", false);
Driver drv2(&config, &mnistLabeled, &mnistcf, &lbfgs2);
drv2.drive();
}
