Fig. 1

Schematic showing how data matrix \(\mathbf {D}\) is decomposed into loading matrix \(\mathbf {P}\) and score matrix \(\mathbf {T}\) in PCA. A given column of \(\mathbf {T}\) and a row of \(\mathbf {P}^{\mathrm{T}}\) form a principal component. The components are sorted such as the variances of the data points in columns of \(\mathbf {T}\) decreases from left to right. The principal components that will be retained after the truncation are filled green