project1.m

模式识别一份很好的作业,包括线性分类器；最小风险贝叶斯分类器；监督学习法分层聚类分析；K－L变换提取有效特征,支持向量机

load vehicle;
load background;
for j=1:10
k_s=1;
k_d=252;
vec_tr(250,252)=0;
vec_test(250,252)=0;
[vec_tr,vec_test]=randQ(vehicle);
randQ(background);
bg_tr(250,252)=0;
bg_test(250,252)=0;
[bg_tr,bg_test]=randQ(background);
M_vec=vec_tr(1,k_s:k_d);
for i=2:250
M_vec=M_vec+(vec_tr(i,k_s:k_d));
end
M_vec=M_vec/250;
M_bg=bg_tr(1,k_s:k_d);
for i=2:250
M_bg=M_bg+(bg_tr(i,k_s:k_d));
end
M_bg=M_bg/250;
d_vec=(M_vec*M_vec')/2;
d_bg=(M_bg*M_bg')/2;
flag_vec(250)=0;
flag_bg(250)=0;
dk_vec=0;
dk_bg=0;
for i=1:250
    dk_vec=vec_test(i,k_s:k_d)*M_vec'-d_vec;
    dk_bg=vec_test(i,k_s:k_d)*M_bg'-d_bg;
    if(dk_vec>=dk_bg)
        flag_vec(i)=1;
    else flag_vec(i)=0;
    end
end
s1_vec=sum(flag_vec);
for i=1:250
    dk_vec=bg_test(i,k_s:k_d)*M_vec'-d_vec;
    dk_bg=bg_test(i,k_s:k_d)*M_bg'-d_bg;
    if(dk_bg>=dk_vec)
        flag_bg(i)=1;
    else flag_bg(i)=0;
    end
end
s1_bg=sum(flag_bg);
err_vg(j)=(250-s1_vec)/250;
err_gv(j)=(250-s1_bg)/250;
end
err_VG=0;err_GV=0;
for i=1:10
    err_VG=err_VG+err_vg(i);
    err_GV=err_GV+err_gv(i);
end
err_VG=err_VG/10
err_GV=err_GV/10