12345qiupeng
/
gtsam
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

add logNormalizingConstant test for GaussianConditional and make some doc fixes

release/4.3a0
Varun Agrawal 2022-12-31 14:43:48 +01:00
parent 878eeb5d73
commit d0a56da726
2 changed files with 27 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
gtsam/linear/GaussianConditional.cpp
View File

@ -67,7 +67,7 @@ namespace gtsam {
GaussianConditional GaussianConditional::FromMeanAndStddev(Key key,
const Vector& mu,
double sigma) {
// |Rx - d| = |x-(Ay + b)|/sigma
// |Rx - d| = |x - mu|/sigma
const Matrix R = Matrix::Identity(mu.size(), mu.size());
const Vector& d = mu;
return GaussianConditional(key, d, R,
@ -189,7 +189,7 @@ double GaussianConditional::logNormalizationConstant() const {
/* ************************************************************************* */
// density = k exp(-error(x))
// log = log(k) -error(x) - 0.5 * n*log(2*pi)
// log = log(k) -error(x)
double GaussianConditional::logDensity(const VectorValues& x) const {
return logNormalizationConstant() - error(x);
}

25
gtsam/linear/tests/testGaussianConditional.cpp
View File

@ -466,6 +466,31 @@ TEST(GaussianConditional, sample) {
// EXPECT(assert_equal(Vector2(31.0111856, 64.9850775), actual2[X(0)], 1e-5));
}
/* ************************************************************************* */
TEST(GaussianConditional, LogNormalizationConstant) {
// Create univariate standard gaussian conditional
auto std_gaussian =
GaussianConditional::FromMeanAndStddev(X(0), Vector1::Zero(), 1.0);
VectorValues values;
values.insert(X(0), Vector1::Zero());
double logDensity = std_gaussian.logDensity(values);
// Regression.
// These values were computed by hand for a univariate standard gaussian.
EXPECT_DOUBLES_EQUAL(-0.9189385332046727, logDensity, 1e-9);
EXPECT_DOUBLES_EQUAL(0.3989422804014327, exp(logDensity), 1e-9);
// Similar test for multivariate gaussian but with sigma 2.0
double sigma = 2.0;
auto conditional = GaussianConditional::FromMeanAndStddev(X(0), Vector3::Zero(), sigma);
VectorValues x;
x.insert(X(0), Vector3::Zero());
Matrix3 Sigma = I_3x3 * sigma * sigma;
double expectedLogNormalizingConstant = log(1 / sqrt((2 * M_PI * Sigma).determinant()));
EXPECT_DOUBLES_EQUAL(expectedLogNormalizingConstant, conditional.logNormalizationConstant(), 1e-9);
}
/* ************************************************************************* */
TEST(GaussianConditional, Print) {
Matrix A1 = (Matrix(2, 2) << 1., 2., 3., 4.).finished();