pcaTransformInv#

Purpose#

Transforms a matrix back to the original feature space of the X which was input to pcaFit().

Format#

X_original = pcaTransformInv(X_transform, mdl)#
Parameters:

  • X_transform (Nxn_components matrix) – The matrix to transform to the original feature space. This matrix must have the same number of columns as the number of principal component vectors in mdl.

  • mdl (pcaModel structure) – This must have been previously been filled in by a call to pcaFit().

Returns:

X_original (NxP matrix) – The input matrix projected back to the original feature space of the matrix input to pcaFit().

Examples#

new;
library gml;

// Get file name with full path
fname = getGAUSSHome("pkgs/gml/examples/winequality.csv");

// Load data
X = loadd(fname, ". -quality");

X_train = X[1:1000,.];
X_test = X[1001:rows(X), .];

n_components = 3;

struct pcaModel mdl;
mdl = pcaFit(X_train, n_components);

// Apply dimensionality reduction to 'X_test'
X_transform = pcaTransform(X_test, mdl);

After the above code, the first 5 rows of X_transform will be:

       PC1              PC2              PC3
 37.441282        1.2145282       -1.5416867
-2.0454164       -15.738950        1.0084994
 21.315231       -2.4328631       0.15655108
 41.776957        2.2901582       -2.2804431
0.73984770       -12.260074      -0.68265628

Now we can use pcaTransformInv() to transform X_transform back to the original space of X, X_train and X_test.

// Transform 'X_transform' back to the original feature space
X_original = pcaTransformInv(X_transform, mdl);

After the above code, the first 5 rows of X_original will be:

7.601  0.548  0.184  2.055  0.088  6.025 11.990  0.996  3.385  0.628 10.346
9.593  0.464  0.343  2.918  0.086 30.986 46.993  0.998  3.290  0.699 10.615
9.071  0.505  0.305  2.526  0.089 13.022 26.985  0.997  3.297  0.667 10.418
6.932  0.568  0.130  1.866  0.087  4.050  7.976  0.996  3.425  0.613 10.312
8.000  0.515  0.221  2.514  0.086 26.987 44.997  0.997  3.379  0.664 10.474

See also

pcaFit(), pcaTransform()

On this page