Large Margin and Local Structure Preservation Sparse Representation Classifier for Alzheimer’s Magnetic Resonance Imaging Classification

. 2022 May 25;14:916020. doi: 10.3389/fnagi.2022.916020

The optimization steps of LMLS-SRC algorithm are shown in Algorithm 1.
Input: training set X and its label matrix Y, tolerance error δ, maximum number of iterations maxiter, parameters λ₁,λ₂,λ₃,λ₄, and λ₅, Output: parameters D, A, and W. Initialize: initialize D and A using the LC-KSVD algorithm, W = I, m = 1, Calculate matrices Q^w, Q^b, and P; While not converged and m≤maxiterdo Calculate D(m) by Eq. (20); Calculate A(m) by Eq. (16); Calculate W(m) by Eq. (22); Check the convergence condition $\frac{\| F (D (m), A (m), W (m)) - F (D (m - 1), A (m - 1), W (m - 1)) \|}{F (D (m - 1), A (m - 1), W (m - 1))} < δ$ m = m + 1 end while

The optimization steps of LMLS-SRC algorithm are shown in Algorithm 1.

Input: training set X and its label matrix Y, tolerance error δ, maximum number of iterations maxiter, parameters λ₁,λ₂,λ₃,λ₄, and λ₅,
Output: parameters D, A, and W.
Initialize: initialize D and A using the LC-KSVD algorithm, W = I, m = 1,
Calculate matrices Q^w, Q^b, and P;
While not converged and m≤maxiterdo
Calculate D(m) by Eq. (20);
Calculate A(m) by Eq. (16);
Calculate W(m) by Eq. (22);
Check the convergence condition

\frac{| F (D (m), A (m), W (m)) - F (D (m - 1), A (m - 1), W (m - 1)) |}{F (D (m - 1), A (m - 1), W (m - 1))} < δ

m = m + 1
end while