stepan
/
dynare
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Estimation DLL: refactor detrending stuff to avoid allocating the matrix for detrended data at every iteration

time-shift
Sébastien Villemot 2010-06-11 19:11:27 +02:00
parent e347f2ae7a
commit e5d093c6ad
10 changed files with 33 additions and 27 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
mex/sources/estimation/DetrendData.cc
View File

@ -31,9 +31,9 @@ DetrendData::DetrendData(const bool INlogLinear) //, Vector& INtrendCoeff)
};
void
DetrendData::detrend(const VectorView &SteadyState, const MatrixConstView &dataView, Matrix &Y)
DetrendData::detrend(const VectorView &SteadyState, const MatrixConstView &dataView,
MatrixView &detrendedDataView)
{
// dummy
Y=dataView;
detrendedDataView = dataView;
};

2
mex/sources/estimation/DetrendData.hh
View File

@ -33,7 +33,7 @@ class DetrendData
public:
virtual ~DetrendData(){};
DetrendData(const bool logLinear); // add later Vector& trendCoeff);
void detrend(const VectorView &SteadyState, const MatrixConstView &dataView, Matrix &Y);
void detrend(const VectorView &SteadyState, const MatrixConstView &dataView, MatrixView &detrendedDataView);
private:
const bool logLinear;

10
mex/sources/estimation/InitializeKalmanFilter.cc
View File

@ -59,10 +59,11 @@ InitializeKalmanFilter::InitializeKalmanFilter(const std::string &dynamicDllFile
void
InitializeKalmanFilter::initialize(VectorView &steadyState, const Vector &deepParams, Matrix &R,
const Matrix &Q, Matrix &RQRt, Matrix &T,
double &penalty, const MatrixConstView &dataView, Matrix &Y, int &info)
double &penalty, const MatrixConstView &dataView,
MatrixView &detrendedDataView, int &info)
{
modelSolution.compute(steadyState, deepParams, g_x, g_u);
detrendData.detrend(steadyState, dataView, Y);
detrendData.detrend(steadyState, dataView, detrendedDataView);
setT(T, info);
setRQR(R, Q, RQRt, info);
@ -72,9 +73,10 @@ InitializeKalmanFilter::initialize(VectorView &steadyState, const Vector &deepPa
void
InitializeKalmanFilter::initialize(VectorView &steadyState, const Vector &deepParams, Matrix &R,
const Matrix &Q, Matrix &RQRt, Matrix &T, Matrix &Pstar, Matrix &Pinf,
double &penalty, const MatrixConstView &dataView, Matrix &Y, int &info)
double &penalty, const MatrixConstView &dataView,
MatrixView &detrendedDataView, int &info)
{
initialize(steadyState, deepParams, R, Q, RQRt, T, penalty, dataView, Y, info);
initialize(steadyState, deepParams, R, Q, RQRt, T, penalty, dataView, detrendedDataView, info);
setPstar(Pstar, Pinf, T, RQRt, info);
}

4
mex/sources/estimation/InitializeKalmanFilter.hh
View File

@ -50,10 +50,10 @@ public:
virtual ~InitializeKalmanFilter();
// initialise all KF matrices
void initialize(VectorView &steadyState, const Vector &deepParams, Matrix &R, const Matrix &Q, Matrix &RQRt,
Matrix &T, Matrix &Pstar, Matrix &Pinf, double &penalty, const MatrixConstView &dataView, Matrix &Y, int &info);
Matrix &T, Matrix &Pstar, Matrix &Pinf, double &penalty, const MatrixConstView &dataView, MatrixView &detrendedDataView, int &info);
// initialise parameter dependent KF matrices only but not Ps
void initialize(VectorView &steadyState, const Vector &deepParams, Matrix &R, const Matrix &Q, Matrix &RQRt,
Matrix &T, double &penalty, const MatrixConstView &dataView, Matrix &Y, int &info);
Matrix &T, double &penalty, const MatrixConstView &dataView, MatrixView &detrendedDataView, int &info);
private:
const double lyapunov_tol;

17
mex/sources/estimation/KalmanFilter.cc
View File

@ -73,18 +73,17 @@ KalmanFilter::compute_zeta_varobs_back_mixed(const std::vector<size_t> &zeta_bac
double
KalmanFilter::compute(const MatrixConstView &dataView, VectorView &steadyState,
const Matrix &Q, const Matrix &H, const Vector &deepParams,
VectorView &vll, size_t start, size_t period, double &penalty, int &info)
VectorView &vll, MatrixView &detrendedDataView,
size_t start, size_t period, double &penalty, int &info)
{
Matrix Y(dataView.getRows(), dataView.getCols()); // data
if(period==0) // initialise all KF matrices
initKalmanFilter.initialize(steadyState, deepParams, R, Q, RQRt, T, Pstar, Pinf,
penalty, dataView, Y, info);
penalty, dataView, detrendedDataView, info);
else // initialise parameter dependent KF matrices only but not Ps
initKalmanFilter.initialize(steadyState, deepParams, R, Q, RQRt, T,
penalty, dataView, Y, info);
penalty, dataView, detrendedDataView, info);
return filter(Y, H, vll, start, info);
return filter(detrendedDataView, H, vll, start, info);
};
@ -92,13 +91,13 @@ KalmanFilter::compute(const MatrixConstView &dataView, VectorView &steadyState,
* 30:*
*/
double
KalmanFilter::filter(const Matrix &dataView, const Matrix &H, VectorView &vll, size_t start, int &info)
KalmanFilter::filter(const MatrixView &detrendedDataView, const Matrix &H, VectorView &vll, size_t start, int &info)
{
double loglik=0.0, ll, logFdet, Fdet;
size_t p = Finv.getRows();
bool nonstationary = true;
for (size_t t = 0; t < dataView.getCols(); ++t)
for (size_t t = 0; t < detrendedDataView.getCols(); ++t)
{
if (nonstationary)
{
@ -138,7 +137,7 @@ KalmanFilter::filter(const Matrix &dataView, const Matrix &H, VectorView &vll,
}
// err= Yt - Za
MatrixConstView yt(dataView, 0, t, dataView.getRows(), 1); // current observation vector
MatrixConstView yt(detrendedDataView, 0, t, detrendedDataView.getRows(), 1); // current observation vector
vt = yt;
blas::gemm("N", "N", -1.0, Z, a_init, 1.0, vt);
// at+1= T(at+ KFinv *err)

5
mex/sources/estimation/KalmanFilter.hh
View File

@ -55,7 +55,8 @@ public:
double compute(const MatrixConstView &dataView, VectorView &steadyState,
const Matrix &Q, const Matrix &H, const Vector &deepParams,
VectorView &vll, size_t start, size_t period, double &penalty, int &info);
VectorView &vll, MatrixView &detrendedDataView, size_t start, size_t period,
double &penalty, int &info);
private:
const std::vector<size_t> zeta_varobs_back_mixed;
@ -77,7 +78,7 @@ private:
InitializeKalmanFilter initKalmanFilter; //Initialise KF matrices
// Method
double filter(const Matrix &data, const Matrix &H, VectorView &vll, size_t start, int &info);
double filter(const MatrixView &detrendedDataView, const Matrix &H, VectorView &vll, size_t start, int &info);
};

9
mex/sources/estimation/LogLikelihoodMain.cc
View File

@ -34,8 +34,8 @@ LogLikelihoodMain::LogLikelihoodMain( //const Matrix &data_arg, Vector &deepPara
: estSubsamples(estiParDesc.estSubsamples),
logLikelihoodSubSample(dynamicDllFile, estiParDesc, n_endo, n_exo, zeta_fwrd_arg, zeta_back_arg, zeta_mixed_arg, zeta_static_arg, qz_criterium,
varobs, riccati_tol, lyapunov_tol, info_arg),
vll(estiParDesc.getNumberOfPeriods()) // time dimension size of data
vll(estiParDesc.getNumberOfPeriods()), // time dimension size of data
detrendedData(varobs.size(), estiParDesc.getNumberOfPeriods())
{
}
@ -55,9 +55,12 @@ LogLikelihoodMain::compute(Matrix &steadyState, const Vector &estParams, Vector
MatrixConstView dataView(data, 0, estSubsamples[i].startPeriod,
data.getRows(), estSubsamples[i].endPeriod-estSubsamples[i].startPeriod+1);
MatrixView detrendedDataView(detrendedData, 0, estSubsamples[i].startPeriod,
data.getRows(), estSubsamples[i].endPeriod-estSubsamples[i].startPeriod+1);
VectorView vllView(vll, estSubsamples[i].startPeriod, estSubsamples[i].endPeriod-estSubsamples[i].startPeriod+1);
logLikelihood += logLikelihoodSubSample.compute(vSteadyState, dataView, estParams, deepParams,
Q, H, vllView, info, start, i);
Q, H, vllView, detrendedDataView, info, start, i);
}
return logLikelihood;
};

1
mex/sources/estimation/LogLikelihoodMain.hh
View File

@ -32,6 +32,7 @@ private:
std::vector<EstimationSubsample> &estSubsamples; // reference to member of EstimatedParametersDescription
LogLikelihoodSubSample logLikelihoodSubSample;
Vector vll; // vector of all KF step likelihoods
Matrix detrendedData;
public:
virtual ~LogLikelihoodMain();

4
mex/sources/estimation/LogLikelihoodSubSample.cc
View File

@ -44,12 +44,12 @@ LogLikelihoodSubSample::LogLikelihoodSubSample(const std::string &dynamicDllFile
double
LogLikelihoodSubSample::compute(VectorView &steadyState, const MatrixConstView &dataView, const Vector &estParams, Vector &deepParams,
Matrix &Q, Matrix &H, VectorView &vll, int &info, size_t start, size_t period)
Matrix &Q, Matrix &H, VectorView &vll, MatrixView &detrendedDataView, int &info, size_t start, size_t period)
{
updateParams(estParams, deepParams, Q, H, period);
if (info == 0)
logLikelihood = kalmanFilter.compute(dataView, steadyState, Q, H, deepParams, vll, start, period, penalty, info);
logLikelihood = kalmanFilter.compute(dataView, steadyState, Q, H, deepParams, vll, detrendedDataView, start, period, penalty, info);
// else
// logLikelihood+=penalty;

2
mex/sources/estimation/LogLikelihoodSubSample.hh
View File

@ -39,7 +39,7 @@ public:
const std::vector<size_t> &varobs_arg, double riccati_tol_in, double lyapunov_tol, int &info);
double compute(VectorView &steadyState, const MatrixConstView &dataView, const Vector &estParams, Vector &deepParams,
Matrix &Q, Matrix &H, VectorView &vll, int &info, size_t start, size_t period);
Matrix &Q, Matrix &H, VectorView &vll, MatrixView &detrendedDataView, int &info, size_t start, size_t period);
virtual ~LogLikelihoodSubSample();
private: