/*
* Analyse the descriptor and any <repository>s that might be part of it
*/
private void analyze(Repository repository) throws Exception {
info("== Analyzing feature descriptor ==");
info(" - read %s", file.getAbsolutePath());
features.add(repository.getFeatures());
// add the repositories from the plugin configuration
if (repositories != null) {
for (String uri : repositories) {
getLog().info(String.format(" - adding repository from %s", uri));
Repository dependency = new RepositoryImpl(URI.create(translateFromMaven(uri)));
schemaCheck(dependency, URI.create(uri));
features.add(dependency.getFeatures());
validateBundlesAvailable(dependency);
analyzeExports(dependency);
}
}
for (URI uri : repository.getRepositories()) {
Artifact artifact = (Artifact) resolve(uri.toString());
Repository dependency =
new RepositoryImpl(new File(localRepo.getBasedir(), localRepo.pathOf(artifact)).toURI());
schemaCheck(dependency, uri);
getLog()
.info(
String.format(
" - adding %d known features from %s", dependency.getFeatures().length, uri));
features.add(dependency.getFeatures());
// we need to do this to get all the information ready for further processing
validateBundlesAvailable(dependency);
analyzeExports(dependency);
}
}
@Test
public void testProcessDep_withConditionalizedAlias() throws Exception {
DependencyList depList =
new DependencyList("test", new HashSet<String>(), mockAggregator, features, true, false);
configRef.set(
new ConfigImpl(
mockAggregator,
tmpDir,
"{aliases:[[/^foo\\//, function(s){return (has('test')?'xxx':'yyy')+'/'}]]}"));
depList.processDep("foo/test", explicitDeps, null, new HashSet<String>(), null);
assertEquals(
new HashSet<String>(Arrays.asList(new String[] {"yyy/test"})), explicitDeps.getModuleIds());
assertEquals(
new HashSet<String>(Arrays.asList(new String[] {"test"})), depList.getDependentFeatures());
depList.getDependentFeatures().clear();
explicitDeps = new ModuleDeps();
features.put("test", true);
depList.processDep("foo/test", explicitDeps, null, new HashSet<String>(), null);
assertEquals(
new HashSet<String>(Arrays.asList(new String[] {"xxx/test"})), explicitDeps.getModuleIds());
assertEquals(
new HashSet<String>(Arrays.asList(new String[] {"test"})), depList.getDependentFeatures());
depList.getDependentFeatures().clear();
explicitDeps = new ModuleDeps();
features.put("test", false);
depList.processDep("foo/test", explicitDeps, null, new HashSet<String>(), null);
assertEquals(
new HashSet<String>(Arrays.asList(new String[] {"yyy/test"})), explicitDeps.getModuleIds());
assertEquals(
new HashSet<String>(Arrays.asList(new String[] {"test"})), depList.getDependentFeatures());
}
private void evaluateAll(
String pluginName,
Features features,
Set<String> discovered,
BooleanTerm term,
ModuleDeps results,
String comment) {
if (feature != null && discovered != null) {
discovered.add(feature);
}
if (feature == null) {
if (nodeName != null && nodeName.length() > 0) {
results.add(nodeName, new ModuleDepInfo(pluginName, term, comment, true));
}
} else if (!features.contains(feature)) {
BooleanTerm newTerm = term.andWith(new BooleanTerm(new BooleanVar(feature, true)));
trueNode.evaluateAll(pluginName, features, discovered, newTerm, results, comment);
newTerm = term.andWith(new BooleanTerm(new BooleanVar(feature, false)));
falseNode.evaluateAll(pluginName, features, discovered, newTerm, results, comment);
} else {
if (features.isFeature(feature)) {
trueNode.evaluateAll(pluginName, features, discovered, term, results, comment);
} else {
falseNode.evaluateAll(pluginName, features, discovered, term, results, comment);
}
}
}
@Test
public void testProcessDep_withCompoundHasExpressionAndDefinedFeatures() throws Exception {
// Defined features should have no impact on results.
features.put("test", true);
features.put("test1", false);
features.put("test2", false);
DependencyList depList =
new DependencyList("test", new HashSet<String>(), mockAggregator, features, true, false);
depList.processDep(
"dojo/has!test?test1?foo1/test:bar1/test:test2:foo2/test:bar2/test",
explicitDeps,
null,
new HashSet<String>(),
null);
assertEquals(5, explicitDeps.size());
assertEquals(
new HashSet<String>(
Arrays.asList(
new String[] {
"dojo/has",
"dojo/has!test?test1?foo1/test",
"dojo/has!test?test1?:bar1/test",
"dojo/has!test?:test2?foo2/test",
"dojo/has!test?:test2?:bar2/test"
})),
explicitDeps.getModuleIds());
assertEquals(
new HashSet<String>(Arrays.asList(new String[] {"test", "test1", "test2"})),
depList.getDependentFeatures());
}
public void train(int maxIter) {
features.setEmpiricalWeight(sampleList.size());
for (int i = 0; i < maxIter; ++i) {
features.updateModelWeight(sampleList);
if (features.isConvergence(0.435)) {
break;
}
}
}
private double euclideanDistance(Features point, Centroid centroid) {
if (point.getNumberOfDim() != centroid.getNumberOfDim()) throw new IllegalArgumentException();
double sumOfSquares = 0;
for (int i = 0; i < point.getNumberOfDim(); i++) {
sumOfSquares +=
(point.getDim(i) - centroid.getDim(i)) * (point.getDim(i) - centroid.getDim(i));
}
return Math.sqrt(sumOfSquares);
}
public void add() {
if (digitEditText.getText().toString().equals("")) {
showAlertDialog("Bład", "Nie podano cyfry do dodania");
} else if (drawView.getPoints().size() == 0) {
showAlertDialog("Bład", "Obrazek nie został narysowany");
} else {
Features features = new Features(drawView.getPoints(), drawView.getStrokes());
List<Double> featureValues = features.calculateFeatures();
drawView.cleanDrawing();
writeToFile(featureValues);
}
}
/** 通过SVD定理，计算矩阵的特征向量 传入形参为 <方法：计算协方差矩阵> 的返回值 */
public static double[][] calFeatureVector(double[][] vec) {
Matrix vecMat = new Matrix(vec);
SingularValueDecomposition s = vecMat.svd();
double[][] featureVector = s.getV().getArray();
Features.getInstance().setFeatureVector(featureVector);
return featureVector;
}
/**
* Recursively resolves the current node with the specified features. Any conditionals for
* features contained in <code>features</code> will be replaced with the result of the evaluation.
*
* @param features The features passed to the aggregator.
* @param discovered A map in which all features encountered in the evaluation will be placed.
* This will not necessarily contain all features in the dependency expression. Only the ones
* in the evaluation chain will be included.
* @param coerceUndefinedToFalse If true, then a feature not being defined will be treated the
* same as if the feature were defined with a value of false.
* @return this node
*/
public HasNode resolve(
Features features, Set<String> discovered, boolean coerceUndefinedToFalse) {
if (feature != null && discovered != null) {
discovered.add(feature);
}
if (feature != null && (features.contains(feature) || coerceUndefinedToFalse)) {
replaceWith(features.isFeature(feature) ? trueNode : falseNode);
resolve(features, discovered, coerceUndefinedToFalse);
}
if (trueNode != null) {
trueNode.resolve(features, discovered, coerceUndefinedToFalse);
}
if (falseNode != null) {
falseNode.resolve(features, discovered, coerceUndefinedToFalse);
}
return this;
}
/**
* Evaluate the has plugin for the given set of features.
*
* @param features The features passed to the aggregator.
* @param discovered A map in which all features encountered in the evaluation will be placed.
* This will not necessarily contain all features in the dependency expression. Only the ones
* in the evaluation chain will be included.
* @param coerceUndefinedToFalse If true, then a feature not being defined will be treated the
* same as if the feature were defined with a value of false.
* @return The evaluated resource based on provided features. If a lack of features prevents us
* from being able to determine the resource, then null is returned. If the required features
* are provided but the evaluation results in no module name, then the empty string is
* returned.
*/
public String evaluate(
Features features, Set<String> discovered, boolean coerceUndefinedToFalse) {
if (feature != null && discovered != null) {
discovered.add(feature);
}
if (feature == null) {
return nodeName;
} else if (!coerceUndefinedToFalse && !features.contains(feature)) {
return null;
} else {
if (features.isFeature(feature)) {
return trueNode.evaluate(features, discovered, coerceUndefinedToFalse);
} else {
return falseNode.evaluate(features, discovered, coerceUndefinedToFalse);
}
}
}
public void loadData(String path) throws IOException {
BufferedReader br = new BufferedReader(new FileReader(new File(path)));
String line = br.readLine();
while (line != null) {
String[] segs = line.split("\\s");
String label = segs[0];
List<String> featureList = new ArrayList<String>();
for (int i = 1; i < segs.length; ++i) {
featureList.add(segs[i]);
features.addFeature(label, segs[i]);
}
features.setMaxFeaturePerSample(featureList.size());
Sample sample = new Sample(label, featureList);
sampleList.add(sample);
labelSet.add(label);
line = br.readLine();
}
}
/**
* 将X'T*X'的特征向量 转换成 X'*X'T的特征向量 的方法
*
* @param eigenValue为特征值
* @param featureVec为特征向量
*/
public static double[][] changeFeatureVector() {
double[][] AveDeviation = Features.getInstance().getAveDev(); // 每幅图的均差
double[] eigenValue = Features.getInstance().getEigenValue(); // 特征值
double[][] featureVec = Features.getInstance().getFeatureVector(); // 特征向量
// System.out.println("width*height:" + width*height);//4096
// System.out.println("featureVec[0].length" + featureVec[0].length);//10
double[][] resultFeatureVector = new double[width * height][featureVec[0].length];
for (int i = 0; i < featureVec[0].length; i++) {
// 将N副图片的特征向量[N][N]分离提取出来，提取第i张图片的特征向量[N][1]
double[][] temp = new double[featureVec.length][1];
for (int j = 0; j < featureVec.length; j++) {
temp[j][0] = featureVec[j][i];
}
Matrix xOld = new Matrix(AveDeviation);
Matrix featureVecOfOneObject = new Matrix(temp);
// System.out.println("[" + xOld.getRowDimension() + "]" +
// "[" + xOld.getColumnDimension() +"]");//[16384][10]
// System.out.println("[" + featureVecOfOneObject.getRowDimension() + "]" +
// "[" + featureVecOfOneObject.getColumnDimension() +"]");//[10][1]
Matrix resultMatrix = xOld.times(featureVecOfOneObject);
// System.out.println("[" + resultMatrix.getRowDimension() + "]" +
// "[" + resultMatrix.getColumnDimension() +"]");
double factor = 1 / (Math.sqrt(eigenValue[i]));
resultMatrix.times(factor);
double[][] temp2 = new double[1][1];
temp2 = resultMatrix.getArray();
// System.out.println("temp2.length:" + temp2.length);//16384-------------
// System.out.println("temp2[0].length:" + temp2[0].length);//1
// 将构成的[N][1] 复制到[N][i]中
for (int j = 0; j < temp2.length; j++) {
resultFeatureVector[j][i] = temp2[j][0];
}
}
Features.getInstance().setResultFeatureVector(resultFeatureVector);
return resultFeatureVector;
}
/** 通过SVD定理，计算矩阵的特征值 传入形参为 <方法：计算协方差矩阵> 的返回值 */
public static double[] calEigenValue(double[][] vec) {
// double[] eigenValue;
Matrix vecMat = new Matrix(vec);
SingularValueDecomposition s = vecMat.svd();
Matrix svalues = new Matrix(s.getSingularValues(), 1);
double[] result = svalues.getRowPackedCopy();
Features.getInstance().setEigenValue(result);
return result;
}
@Test
public void testProcessDep_withAliasResolutionUsingDefinedFeatures1() throws Exception {
DependencyList depList =
new DependencyList("test", new HashSet<String>(), mockAggregator, features, true, false);
features.put("testFoo", true);
features.put("testBar", false);
configRef.set(
new ConfigImpl(
mockAggregator,
tmpDir,
"{aliases:[[/^foo\\//, function(s){return (has('testFoo')?'xxx':'yyy')+'/'}],[/^bar\\//, function(s){return (has('testBar')?'www':'zzz')+'/'}]]}"));
depList.processDep("bar/plugin!foo/test", explicitDeps, null, new HashSet<String>(), null);
assertEquals(
new HashSet<String>(Arrays.asList(new String[] {"zzz/plugin!xxx/test", "zzz/plugin"})),
explicitDeps.getModuleIds());
assertEquals(
new HashSet<String>(Arrays.asList(new String[] {"testFoo", "testBar"})),
depList.getDependentFeatures());
}
public static void main(String argv[]) {
Features.init_shogun_with_defaults();
DoubleMatrix traindata_real = Load.load_numbers("../data/fm_train_real.dat");
DoubleMatrix testdata_real = Load.load_numbers("../data/fm_test_real.dat");
RealFeatures feats_train = new RealFeatures(traindata_real);
RealFeatures feats_test = new RealFeatures(testdata_real);
TanimotoDistance distance = new TanimotoDistance(feats_train, feats_train);
DoubleMatrix dm_train = distance.get_distance_matrix();
distance.init(feats_train, feats_test);
DoubleMatrix dm_test = distance.get_distance_matrix();
System.out.println(dm_train.toString());
System.out.println(dm_test.toString());
Features.exit_shogun();
}
public void hasBranchingDisabledTests_withDefinedFeatures() throws Exception {
mockAggregator.getOptions().setOption(IOptions.DISABLE_HASPLUGINBRANCHING, true);
DependencyList depList =
new DependencyList("test", new HashSet<String>(), mockAggregator, features, true, false);
features.put("test", true);
features.put("test1", true);
depList.processDep(
"dojo/has!test?test1?foo1/test:bar1/test:test2:foo2/test:bar2/test",
explicitDeps,
null,
new HashSet<String>(),
null);
assertEquals(2, explicitDeps.size());
assertEquals(
new HashSet<String>(Arrays.asList(new String[] {"dojo/has", "foo1/test"})),
explicitDeps.getModuleIds());
assertEquals(
new HashSet<String>(Arrays.asList(new String[] {"test", "test1"})),
depList.getDependentFeatures());
}
/** 计算一张图片的均差，仅针对一个人的图像做计算 形参vec中存放的数据为1副人脸图像（一维） */
public static double[] calAveDeviationOneObject(double[] vec) {
double[] newVec = new double[vec.length];
double[] vAve = Features.getInstance().getAveVector(); // 获取平均人脸图像
// 计算这张图片的均差
for (int i = 0; i < vec.length; i++) {
newVec[i] = vec[i] - vAve[i];
}
return newVec;
}
/*
* Validate if all imports for a feature are being matched with exports
*/
private void validateImportsExports(Feature feature) throws Exception {
Map<Clause, String> imports = new HashMap<Clause, String>();
Set<Clause> exports = new HashSet<Clause>();
for (Dependency dependency : feature.getDependencies()) {
if (featureExports.containsKey(dependency.getName())) {
exports.addAll(featureExports.get(dependency.getName()));
} else {
validateImportsExports(features.get(dependency.getName(), dependency.getVersion()));
}
}
for (String bundle : getBundleLocations(feature)) {
Manifest meta = manifests.get(bundles.get(bundle));
exports.addAll(getExports(meta));
for (Clause clause : getMandatoryImports(meta)) {
imports.put(clause, bundle);
}
}
// setting up the set of required imports
Set<Clause> requirements = new HashSet<Clause>();
requirements.addAll(imports.keySet());
// now, let's remove requirements whenever we find a matching export for them
for (Clause element : imports.keySet()) {
if (systemExports.contains(element.getName())) {
debug("%s is resolved by a system bundle export or provided bundle", element);
requirements.remove(element);
continue;
}
for (Clause export : exports) {
if (matches(element, export)) {
debug("%s is resolved by export %s", element, export);
requirements.remove(element);
continue;
}
debug("%s is not resolved by export %s", element, export);
}
}
// if there are any more requirements left here, there's a problem with the feature
if (!requirements.isEmpty()) {
warn("Failed to validate feature %s", feature.getName());
for (Clause entry : requirements) {
warn("No export found to match %s (imported by %s)", entry, imports.get(entry));
}
throw new Exception(
String.format(
"%d unresolved imports in feature %s", requirements.size(), feature.getName()));
}
info(" OK: imports resolved for %s", feature.getName());
featureExports.put(feature.getName(), exports);
}
@Test
public void resolveAliasesDisabledAndHasBranchingDisabled_withDefinedFeatures() throws Exception {
mockAggregator.getOptions().setOption(IOptions.DISABLE_HASPLUGINBRANCHING, true);
configRef.set(new ConfigImpl(mockAggregator, tmpDir, "{aliases:[['foo1/test', 'bar/test']]}"));
DependencyList depList =
new DependencyList("test", new HashSet<String>(), mockAggregator, features, false, false);
features.put("test", true);
features.put("test1", true);
depList.processDep(
"dojo/has!test?test1?foo1/test:bar1/test:test2:foo2/test:bar2/test",
explicitDeps,
null,
new HashSet<String>(),
null);
assertEquals(2, explicitDeps.size());
assertEquals(
new HashSet<String>(Arrays.asList(new String[] {"dojo/has", "foo1/test"})),
explicitDeps.getModuleIds());
assertEquals(
new HashSet<String>(Arrays.asList(new String[] {"test", "test1"})),
depList.getDependentFeatures());
}
/** 计算每张图片的均差 形参vec中存放的数据为N副人脸图像（一维）组成的二维向量 形参vAve为平均人脸图像的向量 */
public static double[][] calAveDeviation(double[][] vec, double[] vAve) {
int length = vec.length; // length指向量的维数，X={1,2,3,...,length}
int n = vec[0].length; // n指向量的个数，X1,X2,X3...Xn;
double[][] newVec = new double[length][n];
// 计算每张图片的均差
for (int i = 0; i < length; i++) {
for (int j = 0; j < n; j++) {
newVec[i][j] = vec[i][j] - vAve[i];
}
}
Features.getInstance().setAveDev(newVec);
return newVec; // newVec为每幅图像的均差，同时也可以表示成一个矩阵。维数为：n*length;
}
/**
* transform the code to a new form.
*
* @param type the new form, one of {@link #TYPE_STACK} or {@link #TYPE_QUAD}.
* @return true if the graph has been transformed, false if the graph was already in the requested
* form.
* @throws GraphException if conversion of statements fails.
*/
public boolean transformTo(int type) throws GraphException {
if (type == graphType || type == TYPE_BOTH) {
return false;
}
for (Iterator it = blocks.iterator(); it.hasNext(); ) {
GraphBlock block = (GraphBlock) it.next();
block.transformTo(type);
}
// TODO any feature which survives transformation? (getter from Feature)
features.clearFeatures();
graphType = type;
return true;
}
// calculate p(y|x)=sum_i(alpha_i*f_i(x,y))/sum_y(sum_i(alpha_i*f_i(x,y))).
public Map<String, Double> predict(List<String> x) {
Map<String, Double> probMap = new HashMap<String, Double>();
double sum = 0;
for (String label : labelSet) {
double prob = 0;
for (String feature : x) {
prob += features.getModelWeight(label, feature);
}
prob = Math.exp(prob);
probMap.put(label, prob);
sum += prob;
}
for (String label : probMap.keySet()) {
probMap.put(label, probMap.get(label) / sum);
}
return probMap;
}
/** 计算平均人脸图像 形参vec中存放的数据为N副人脸图像（一维）合成的二维数据 */
public static double[] calAveVector(double[][] vec) {
int length = vec.length; // length指向量的维数，X={1,2,3,...,length}
int n = vec[0].length; // n指向量的个数，X1,X2,X3...Xn;
double[] vAve = new double[length]; // 初始化平均向量(的维数)
// 构建平均向量的值
for (int i = 0; i < length; i++) {
double temp = 0;
for (int j = 0; j < n; j++) {
temp += vec[i][j];
}
temp = temp / n;
vAve[i] = temp;
}
Features.getInstance().setAveVector(vAve);
return vAve;
}
/**
* 输出特征脸 根据特征向量得出Image类型
*
* @param args
*/
public static Image[] getFeatureFaces() {
double[][] rfVec = Features.getInstance().getResultFeatureVector();
Matrix mat = new Matrix(rfVec);
mat = mat.transpose();
rfVec = mat.getArray();
Image[] img = new Image[rfVec.length];
for (int i = 0; i < rfVec.length; i++) {
System.out.println("i:" + i);
// 这一步可能会有压缩损失
// 最好找一个方法，直接从double[]转换成Image
// 将double[]转换成int[]
int[][] rfVecInt = new int[rfVec.length][rfVec[0].length];
// System.out.println("[" + rfVec.length + "]" + "[" + rfVec[0].length + "]");
for (int j = 0; j < rfVec[0].length; j++) {
// rfVecInt[i][j] = (int)rfVec[i][j];//无四舍五入
rfVecInt[i][j] =
Integer.parseInt(new java.text.DecimalFormat("0").format(rfVec[i][j])); // 四舍五入
// System.out.print(rfVecInt[i][j] + " ");
// ----------------整个if语句是测试语句，要删除
// if(i==9)
// {
// System.out.println("Yes!");
// rfVecInt[i][j] = Integer.parseInt(new
// java.text.DecimalFormat("0").format(rfVec[i][j]));//四舍五入
// }
// ----------------以上为测试语句
}
System.out.println("i:" + i);
img[i] = getImgByPixels(width, height, rfVecInt[i]);
// System.out.println();
// img[i] = getImgByPixels(128, 128, rfVecInt[i]);
}
// System.out.println("img length:" + img.length);
return img;
}
/** 计算协方差矩阵 传入形参为 <方法：计算每张图片的均差> 的返回值 (其实这并不是协方差矩阵，若需要协方差矩阵，再用该矩阵除以N即可) */
public static double[][] calCovarianceMatrix(double[][] vec) {
int length = vec.length; // length指向量的维数，X={1,2,3,...,length}
int n = vec[0].length; // n指向量的个数，X1,X2,X3...Xn;
Matrix vecOld = new Matrix(vec);
// System.out.println("vecOld:");
// vecOld.print(6, 2);
// System.out.println();
Matrix vecTrans = vecOld.transpose(); // 获取转置矩阵
// System.out.println("vecTrans:");
// vecTrans.print(6, 2);
double[][] covArr = new double[nOld][nOld]; // 初始化协方差矩阵
// 构造协方差矩阵
// Matrix cov = vecOld.times(vecTrans);
Matrix cov = vecTrans.times(vecOld);
covArr = cov.getArray();
// 由于最终计算并不需要协方差矩阵，仅仅要的是一部分，所以不用除以N
// 除以N
// for(int i=0; i<n; i++)
// {
// for(int j=0; j<length; j++)
// {
// covArr[i][j] = covArr[i][j] / n;
// }
// }
Features.getInstance().setCovarianceMatrix(covArr);
return covArr;
}
public static void main(String[] args) {
// 创建图片数组
ImageIcon[] icon = new ImageIcon[10];
String source = "Pictures/monkey-test/after";
for (int i = 0; i < 6; i++) {
String target = "-" + (i + 1) + ".jpg";
icon[i] = new ImageIcon(source + target);
}
icon[6] = new ImageIcon(source + "1.jpg");
for (int i = 7; i < 10; i++) {
String target = (i - 4) + ".jpg";
icon[i] = new ImageIcon(source + target);
}
// 计算图片数组的特征值和特征向量
// 将结果保存在Features类中
imageToResult(icon);
// 测试---------------------------
// 将特征向量转换成图片
getFeatureFaces();
// 从单例中获取数据
double[] eigenValue = Features.getInstance().getEigenValue();
double[][] featureVector = Features.getInstance().getFeatureVector();
System.out.println("特征值的个数：" + eigenValue.length);
System.out.println(
"特征向量的维数："
+ "["
+ Features.getInstance().getResultFeatureVector().length
+ "]"
+ "["
+ Features.getInstance().getResultFeatureVector()[0].length
+ "]");
// System.out.println("特征向量的维数：" +
// "[" + featureVector.length +"]" + "[" + featureVector[0].length+ "]");
// 输出测试
System.out.println("特征值：");
for (int i = 0; i < eigenValue.length; i++) {
System.out.print(eigenValue[i] + " ");
}
System.out.println();
System.out.println("---------分割线------------");
// System.out.println("特征向量：");
// double[][] result = Features.getInstance().getResultFeatureVector();
// for(int i=0; i<result.length; i++)
// {
// for(int j=0; j<result[i].length; j++)
// {
// System.out.print(result[i][j] + " ");
// }
// System.out.println();
// }
// System.out.println("特征向量：");
// for(int i=0; i<featureVector.length; i++)
// {
// for(int j=0; j<featureVector[i].length; j++)
// {
// System.out.print(featureVector[i][j] + " ");
// }
// System.out.println();
// }
// //可以将这里封装成一个方法
// //形参为图片数组，返回值为double[][]
// ImageIcon icon = new ImageIcon("Pictures/monkey-test/after-1.jpg");
// icon = ImageHandle(icon, width, height);
// Image img = icon.getImage();
// BufferedImage bimg = ImageUtil.ImageToBufferedImage(img);
// int[] imgVec = getPixes(bimg);
// double[] imgVecD = intToDouble(imgVec);
//
//// System.out.println("length:" + imgVec.length);
//
// //先用一张图片测试，这里实际应该为N张图片
// double[][] imgVecD2 = new double[imgVecD.length][1];
// for(int i=0; i<imgVecD.length; i++)
// {
// imgVecD2[i][0] = imgVecD[i];//构造列向量
// }
//// imgVecD2[0] = imgVecD;
//
// //计算平均人脸图像
// double[] aveVec = calAveVector(imgVecD2);
//
// //计算每张图片的均差
// double[][] aveDev = calAveDeviation(imgVecD2, aveVec);
//
// //计算协方差矩阵
// double[][] covMat = calCovarianceMatrix(aveDev);
//
// //通过SVD定理，计算矩阵的特征值
// double[] eigenValue = calEigenValue(covMat);
//
// //通过SVD定理，计算矩阵的特征向量
// double[][] featureVector = calFeatureVector(covMat);
//
// //输出测试
// System.out.println("特征值：");
// for(int i=0; i<eigenValue.length; i++) {
// System.out.print(eigenValue[i] + " ");
// }
// System.out.println();
// System.out.println("---------分割线------------");
//
// System.out.println("特征向量：");
// for(int i=0; i<featureVector.length; i++)
// {
// for(int j=0; j<featureVector[i].length; j++)
// {
// System.out.print(featureVector[i][j] + " ");
// }
// System.out.println();
// }
//
// ---------------------真.分割线---------------------------------
// double [][] array = { {4,6,0,3},{-3,-5,0,3},{-3,-6,1,3}};
// System.out.println(array.length);
// System.out.println(array[0].length);
//
//
// Matrix a = new Matrix(array);
// double[][] ar = {{3,-2,1},{0,0,0}};
// double[][] ar2 = {{1,0,-1},{0,0,0}};
//
// Matrix a = new Matrix(ar);
// Matrix a2 = new Matrix(ar2);
// a2 = a2.transpose();
//
// Matrix a3 = a.times(a2);
// a3.print(4, 1);
// double[][] ar = a.getArray();
// double[][] ar = a.transpose().getArray();
// System.out.println(ar[0][0] + " " +ar[0][1]+" " +ar[0][2]);
// System.out.println(ar[1][0] + " " +ar[1][1]+" " +ar[1][2]);
// System.out.println(ar[2][0] + " " +ar[2][1]+" " +ar[2][2]);
// System.out.println(ar[3][0] + " " +ar[3][1]+" " +ar[3][2]);
// System.out.println(ar[0][0] + " " +ar[0][1]+" " +ar[0][2]+" " +ar[0][3]);
// System.out.println(ar[1][0] + " " +ar[1][1]+" " +ar[1][2]+" " +ar[1][3]);
// System.out.println(ar[2][0] + " " +ar[2][1]+" " +ar[2][2]+" " +ar[2][3]);
// double[] ar = array[0];
// System.out.println(ar[0] + " " +ar[1]+" " +ar[2]+" " +ar[3]);
// double[] arr = {4,2,0,8};
// double[][] ar = new double[3][4];
// ar[0] = arr;
// System.out.println(ar[0][0] + " " +ar[0][1]+" " +ar[0][2]+" " +ar[0][3]);
// create M-by-N matrix that doesn't have full rank
// int M = 8, N = 5;
// Matrix B = Matrix.random(5, 3);
// Matrix A = Matrix.random(M, N).times(B).times(B.transpose());
// System.out.print("A = ");
// A.print(9, 6);
//
// // compute the singular vallue decomposition
// System.out.println("A = U S V^T");
// System.out.println();
// SingularValueDecomposition s = A.svd();
// System.out.print("U = ");
// Matrix U = s.getU();
// U.print(9, 6);
// System.out.print("Sigma = ");
// Matrix S = s.getS();
// S.print(9, 6);
// System.out.print("V = ");
// Matrix V = s.getV(); //特征向量
// V.print(9, 6);
// System.out.println("rank = " + s.rank());
// System.out.println("condition number = " + s.cond());
// System.out.println("2-norm = " + s.norm2());
//
// // print out singular values
// System.out.print("singular values = ");
// Matrix svalues = new Matrix(s.getSingularValues(), 1); //特征值
// svalues.print(9, 6);
}
