autoPCA TableOfReal_to_PCA (I) {
iam (TableOfReal);
try {
long m = my numberOfRows, n = my numberOfColumns;
if (! TableOfReal_areAllCellsDefined (me, 0, 0, 0, 0)) {
Melder_throw (U"Undefined cells.");
}
if (m < 2) {
Melder_throw (U"There is not enough data to perform a PCA.\nYour table has less than 2 rows.");
}
if (m < n) {
Melder_warning (U"The number of rows in your table is less than the \nnumber of columns. ");
}
if (NUMfrobeniusnorm (m, n, my data) == 0) {
Melder_throw (U"All values in your table are zero.");
}
autoPCA thee = Thing_new (PCA);
autoNUMmatrix<double> a (NUMmatrix_copy (my data, 1, m, 1, n), 1, 1);
thy centroid = NUMvector<double> (1, n);
for (long j = 1; j <= n; j++) {
double colmean = a[1][j];
for (long i = 2; i <= m; i++) {
colmean += a[i][j];
}
colmean /= m;
for (long i = 1; i <= m; i++) {
a[i][j] -= colmean;
}
thy centroid[j] = colmean;
}
Eigen_initFromSquareRoot (thee.peek(), a.peek(), m, n);
thy labels = NUMvector<char32 *> (1, n);
NUMstrings_copyElements (my columnLabels, thy labels, 1, n);
PCA_setNumberOfObservations (thee.peek(), m);
/*
The covariance matrix C = A'A / (N-1). However, we have calculated
the eigenstructure for A'A. This has no consequences for the
eigenvectors, but the eigenvalues have to be divided by (N-1).
*/
for (long i = 1; i <= thy numberOfEigenvalues; i++) {
thy eigenvalues[i] /= (m - 1);
}
return thee;
} catch (MelderError) {
Melder_throw (me, U": PCA not created.");
}
}
static autoPCA NUMdmatrix_to_PCA (double **m, long numberOfRows, long numberOfColumns, bool byColumns) {
try {
if (! NUMdmatrix_hasFiniteElements(m, 1, numberOfRows, 1, numberOfColumns)) {
Melder_throw (U"At least one of the matrix elements is not finite or undefined.");
}
if (NUMfrobeniusnorm (numberOfRows, numberOfColumns, m) == 0.0) {
Melder_throw (U"All values in your table are zero.");
}
autoNUMmatrix<double> mcopy;
long numberOfRows2, numberOfColumns2;
if (byColumns) {
if (numberOfColumns < numberOfRows) {
Melder_warning (U"The number of columns in your table is less than the number of rows. ");
}
numberOfRows2 = numberOfColumns, numberOfColumns2 = numberOfRows;
mcopy.reset (1, numberOfRows2, 1, numberOfColumns2);
for (long i = 1; i <= numberOfRows2; i ++) { // transpose
for (long j = 1; j <= numberOfColumns2; j++) {
mcopy [i] [j] = m [j] [i];
}
}
} else {
if (numberOfRows < numberOfColumns) {
Melder_warning (U"The number of rows in your table is less than the number of columns. ");
}
numberOfRows2 = numberOfRows, numberOfColumns2 = numberOfColumns;
mcopy.reset (1, numberOfRows2, 1, numberOfColumns2);
NUMmatrix_copyElements<double>(m, mcopy.peek(), 1, numberOfRows2, 1, numberOfColumns2);
}
autoPCA thee = Thing_new (PCA);
thy centroid = NUMvector<double> (1, numberOfColumns2);
NUMcentreColumns (mcopy.peek(), 1, numberOfRows2, 1, numberOfColumns2, thy centroid);
Eigen_initFromSquareRoot (thee.get(), mcopy.peek(), numberOfRows2, numberOfColumns2);
thy labels = NUMvector<char32 *> (1, numberOfColumns2);
PCA_setNumberOfObservations (thee.get(), numberOfRows2);
/*
The covariance matrix C = A'A / (N-1). However, we have calculated
the eigenstructure for A'A. This has no consequences for the
eigenvectors, but the eigenvalues have to be divided by (N-1).
*/
for (long i = 1; i <= thy numberOfEigenvalues; i++) {
thy eigenvalues [i] /= (numberOfRows2 - 1);
}
return thee;
} catch (MelderError) {
Melder_throw (U"No PCA created from ", byColumns ? U"columns." : U"rows.");
}
}
