人工智能-matlab-基于类别的稀疏表示分类器的简单实现，MATLAB代码版本

close all; clear; clc; load('Indiana200.mat'); Indiana_data=fea'; Indiana_labels=labels'; % load('Indiana_200.mat'); % data_ori = Indiana_data(find(Indiana_labels ~= 0), :)'; % labels = Indiana_labels(find(Indiana_labels ~= 0), :)'; lyl=[2,3,5,6,8,10,11,12,14]; data_ori =[]; labels=[]; for j=1:9 data_ori =[data_ori,Indiana_data(find(Indiana_labels == lyl(j)), :)' ]; labels = [labels ,repmat(j,1,length(Indiana_labels(find(Indiana_labels == lyl(j)), :)'))]; end %数据归一化 data = normalize_data(data_ori); %对原始数据进行LFDA映射降到30维，用于后面的cdKNN [T,ZZ]=lda(data_ori', labels', 30); %T是映射矩阵，Z是降维后的数据T'*X Z=ZZ'; % [T,Z]=LFDA(data_ori, labels, 30); %T是映射矩阵，Z是降维后的数据T'*X %选取训练样本和测试样本 %select_train_data.m用来按比例选择训练样本 %select_train_data1.m用来按个数选择训练样本 percent = 0.1; %每类样本中训练样本比例 %N = 10; %每类取N个作为训练样本 [train_index, test_index] = select_train_data(labels, 0.1); % [train_index, test_index] = select_train_data1(labels, 30); %用于cdOMP的归一化后数据 train_data = data(:, train_index); train_label = labels(train_index); test_data = data(:, test_index); test_label = labels(test_index); %用于cdKNN的降维数据 train_data_ori = Z(:, train_index); test_data_ori = Z(:, test_index); X = train_data; S = 10; K = 10; lambda = 0.0001; c = max(labels); %cdOMP迭代程序，得到【类别数*测试样本个数】大小的矩阵residual tic residual = zeros(max(train_label), length(test_label)); distance = zeros(size(residual)); for i = 1:max(train_label) X1 = X(:, find(train_label == i)); A= OMP(X1, test_data, S); nor = sqrt(sum((X1 * A - test_data).^2)); residual(i, :) = nor; end %cdKNN程序，得到【类别数*测试样本个数】大小的矩阵distance distance = cdKNN(train_data_ori, test_data_ori, train_label, K); %整合相关度信息和欧氏距离信息 w = residual + lambda * distance; %计算正确率 result = zeros(1, length(test_label)); for i = 1:length(test_label) result(i) = find(w(:, i) == min(w(:, i))); end per = sum(result == test_label)/length(test_label) toc; % tic; % %result = zeros(1, 1000); % w = zeros(1, length(test_label)); %w用来保存新的类标 % for i = 1:length(test_label) % residual = cdOMP(X, test_data(:, i), train_label, S); %差值，y关于每类的相关性度量 % distance = cdKNN(X, test_data(:, i), train_label, K); %距离，y关于每类的欧氏距离性度量 % W = residual + lambda * distance; % w(i) = find(W == min(W)); % end % %a = test_label(1:1000); % per = length(find(test_label == w)) % %percent = lenght(find(w == test_label(1:100)))/100; % toc;