Two-Stage Feature Generator for Handwritten Digit Classification

. 2023 Oct 15;23(20):8477. doi: 10.3390/s23208477

Algorithm 1: Obtaining principal component (PC)-based features

The data matrix D = (d_{1} d_{2} \dots d_{N})

of size M x N

where d_{i}

represents the i^{t h}

sample from the data matrix, where i = 1, \dots, N

and N

is the number of examples in data matrix.

$S 1 : Scale the values of the data matrix in [0, 1]$ $. Resulting matrix is called D_{s}$ .
$S 2 : Calculate the principal components (PCs) of the D_{s}$ .
$S 3 : Select the PCs corresponding to K$ highest eigenvalues.
$S 4 : Construct the matrix whose columns are formed by the principal component vectors (eigenvectors of D_{s}$ $) C = (c_{1} c_{2} \dots c_{K})$ $with the size of M x K .$
S5: Calculate the feature matrix by

$F = D_{s}^{T} C = (d_{1} c_{1} d_{1} c_{2} \dots d_{1} c_{K})$ $with the size of N x K$ .

Although the algorithm ends in this step, the following step demonstrates the effectiveness of the generated features.
$S 6 : Train a classifier (such as SVM or MDC) by the rows of the matrix F$ .