================================================================================ * Feature extraction original_feature_vector=[x_1,x_2,...x_N] dim_reduced_feat_vector=linear_feature_extraction(original_feature_vector) print(dim_reduced_feat_vector) # [y_1,y_2,...y_M] ================================================================================ * $$$X = f(X) = \mathbb{Y}^M = \mathbb{W}^{M\times N} \mathbb{X}^N$$$ * dim_reduced_feat_vector = transform_mat $$$\times$$$ original_feature_vector ================================================================================ * W can be very various based on your purposes which feature you want to more preserve * Purpose1: signal representation - Precise data representaion in lower dimension - W is found by PCA * Purpose2: classification - Easier classification in lower dimension - W is found by LDA ================================================================================ * 2D feature vector * 1D feature vector for better signal representation using PCA - Variance of each feature is preserved * 1D feature vector for better classification using LDA - Easier classification