genTiny.m

%%
clear all;
clc;
path(path,'dataIO');
%%
load data/MH数据库/Tiny100k.mat;   %
A=1:100000;
ntesdex=randperm(100000,10000);
B=setdiff(A,ntesdex);
ranrow=randperm(90000);
ntrndex=B(1,ranrow(1,:));

Xtraining= single(X(ntrndex(1,:),:));
Xtest = single(X(ntesdex(1,:),:));

% load data/cifar_10yunchao.mat  %cifar10 
% load data/cifar-10-batches-mat/test_batch.mat %计算S矩阵用
% Xtraining= single(cifar10(1:59000,1:end-1));
% Xtest = single(cifar10(59001:60000,1:end-1));
% XtestLabels=labels(9001:end,:);
% Scifar=getS(XtestLabels);

MM=mean(Xtraining, 1);
Xtraining=Xtraining-repmat(MM,length(Xtraining),1);
Xtest=Xtest-repmat(MM,size(Xtest,1),1);
%%
figure;
scatter(Xtraining(:,1), Xtraining(:,2), 3, 'b');
hold on;
scatter(Xtest(:,1), Xtest(:,2), 3, 'r');

% define ground-truth neighbors (this is only used for the evaluation):
Nneighbors=0.01*length(Xtraining);
DtrueTestTraining = distMat(Xtest,Xtraining); % size = [Ntest x Ntraining]
%测试样本与训练样本之间的距离
[Dball, I] = sort(DtrueTestTraining,2); %按行排列，每一行表示测试样本数据点与训练样本数据点的距离
KNN_info.knn_p2=I(:,1:Nneighbors); %保存测试数据点的1000个近邻索引 10000*1000
KNN_info.dis_p2=Dball(:,1:Nneighbors); %保存测试数据点的1000个近邻的欧式距离 10000*1000
%save('.\data\Data_sift1M','Xtraining','Xtest','KNN_info');
%save('.\data\Data_cifar','Xtraining','Xtest','KNN_info');
%save('.\data\S_cifar','Scifar');
% save('.\data\Data_MNIST','Xtraining','Xtest','KNN_info');
 save('.\data\Data_Tiny','Xtraining','Xtest','KNN_info');