Algorithm 1. Pseudocode of the NCAR algorithm

Input: feature vector from CovEncoNet model (fea), size of feature vector (N)   average of fea(avg), standard deviation of fea (std), threshold (thr) Output: reduced feture vector (fea_out) 1:. feature_reducion(fea,std,avg,thr) 2:. begin 2:. fea_out = fea 3:. for i = 1 to N do 4:. decision1 = std/fea_out[i] 5:. decision2 = avg/fea_out[i] 6:. if decision1 > thr and decision2 > thr 7:. fea_out[i] = [] 8:. end if 9:. end for i 10:. end