本文整理汇总了Java中org.apache.commons.math.util.MathUtils.EPSILON属性的典型用法代码示例。如果您正苦于以下问题：Java MathUtils.EPSILON属性的具体用法？Java MathUtils.EPSILON怎么用？Java MathUtils.EPSILON使用的例子？那么, 这里精选的属性代码示例或许可以为您提供帮助。您也可以进一步了解该属性所在类org.apache.commons.math.util.MathUtils的用法示例。
在下文中一共展示了MathUtils.EPSILON属性的9个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于系统推荐出更棒的Java代码示例。
示例1: mstep
@Override
protected void mstep(EMGMM gmm, GaussianMixtureModelEM learner, Matrix X, Matrix responsibilities,
Matrix weightedXsum,
double[] norm)
{
// Eq. 12 from K. Murphy,
// "Fitting a Conditional Linear Gaussian Distribution"
final int nfeatures = X.getColumnDimension();
for (int c = 0; c < learner.nComponents; c++) {
final Matrix post = responsibilities.getMatrix(0, X.getRowDimension() - 1, c, c).transpose();
final double factor = 1.0 / (ArrayUtils.sumValues(post.getArray()) + 10 * MathUtils.EPSILON);
final Matrix pXt = X.transpose();
for (int i = 0; i < pXt.getRowDimension(); i++)
for (int j = 0; j < pXt.getColumnDimension(); j++)
pXt.set(i, j, pXt.get(i, j) * post.get(0, j));
final Matrix argcv = pXt.times(X).times(factor);
final Matrix mu = ((FullMultivariateGaussian) gmm.gaussians[c]).mean;
((FullMultivariateGaussian) gmm.gaussians[c]).covar = argcv.minusEquals(mu.transpose().times(mu))
.plusEquals(Matrix.identity(nfeatures, nfeatures).times(learner.minCovar));
}
}
示例2: isSymmetric
/**
* Check if a matrix is symmetric.
* @param matrix matrix to check
* @return true if matrix is symmetric
*/
private boolean isSymmetric(final RealMatrix matrix) {
final int rows = matrix.getRowDimension();
final int columns = matrix.getColumnDimension();
final double eps = 10 * rows * columns * MathUtils.EPSILON;
for (int i = 0; i < rows; ++i) {
for (int j = i + 1; j < columns; ++j) {
final double mij = matrix.getEntry(i, j);
final double mji = matrix.getEntry(j, i);
if (Math.abs(mij - mji) > (Math.max(Math.abs(mij), Math.abs(mji)) * eps)) {
return false;
}
}
}
return true;
}
示例3: isSymmetric
/**
* Check if a matrix is symmetric.
*
* @param matrix Matrix to check.
* @param raiseException If {@code true}, the method will throw an
* exception if {@code matrix} is not symmetric.
* @return {@code true} if {@code matrix} is symmetric.
* @throws NonSymmetricMatrixException if the matrix is not symmetric and
* {@code raiseException} is {@code true}.
*/
private boolean isSymmetric(final RealMatrix matrix,
boolean raiseException) {
final int rows = matrix.getRowDimension();
final int columns = matrix.getColumnDimension();
final double eps = 10 * rows * columns * MathUtils.EPSILON;
for (int i = 0; i < rows; ++i) {
for (int j = i + 1; j < columns; ++j) {
final double mij = matrix.getEntry(i, j);
final double mji = matrix.getEntry(j, i);
if (FastMath.abs(mij - mji) >
(FastMath.max(FastMath.abs(mij), FastMath.abs(mji)) * eps)) {
if (raiseException) {
throw new NonSymmetricMatrixException(i, j, eps);
}
return false;
}
}
}
return true;
}
示例4: isSymmetric
/**
* Check if a matrix is symmetric.
* @param matrix
* matrix to check
* @return true if matrix is symmetric
*/
private boolean isSymmetric(final RealMatrix matrix) {
final int rows = matrix.getRowDimension();
final int columns = matrix.getColumnDimension();
final double eps = 10 * rows * columns * MathUtils.EPSILON;
for (int i = 0; i < rows; ++i) {
for (int j = i + 1; j < columns; ++j) {
final double mij = matrix.getEntry(i, j);
final double mji = matrix.getEntry(j, i);
if (Math.abs(mij - mji) > (Math.max(Math.abs(mij), Math
.abs(mji)) * eps)) {
return false;
}
}
}
return true;
}
示例5: findEigenvalues
/**
* Find the realEigenvalues.
* @exception InvalidMatrixException if a block cannot be diagonalized
*/
private void findEigenvalues()
throws InvalidMatrixException {
// compute splitting points
List<Integer> splitIndices = computeSplits();
// find realEigenvalues in each block
realEigenvalues = new double[main.length];
imagEigenvalues = new double[main.length];
int begin = 0;
for (final int end : splitIndices) {
final int n = end - begin;
switch (n) {
case 1:
// apply dedicated method for dimension 1
process1RowBlock(begin);
break;
case 2:
// apply dedicated method for dimension 2
process2RowsBlock(begin);
break;
case 3:
// apply dedicated method for dimension 3
process3RowsBlock(begin);
break;
default:
// choose an initial shift for LDL<sup>T</sup> decomposition
final double[] range = eigenvaluesRange(begin, n);
final double oneFourth = 0.25 * (3 * range[0] + range[1]);
final int oneFourthCount = countEigenValues(oneFourth, begin, n);
final double threeFourth = 0.25 * (range[0] + 3 * range[1]);
final int threeFourthCount = countEigenValues(threeFourth, begin, n);
final boolean chooseLeft = (oneFourthCount - 1) >= (n - threeFourthCount);
final double lambda = chooseLeft ? range[0] : range[1];
tau = (range[1] - range[0]) * MathUtils.EPSILON * n + 2 * minPivot;
// decompose T-λI as LDL<sup>T</sup>
ldlTDecomposition(lambda, begin, n);
// apply general dqd/dqds method
processGeneralBlock(n);
// extract realEigenvalues
if (chooseLeft) {
for (int i = 0; i < n; ++i) {
realEigenvalues[begin + i] = lambda + work[4 * i];
}
} else {
for (int i = 0; i < n; ++i) {
realEigenvalues[begin + i] = lambda - work[4 * i];
}
}
}
begin = end;
}
// sort the realEigenvalues in decreasing order
Arrays.sort(realEigenvalues);
int j = realEigenvalues.length - 1;
for (int i = 0; i < j; ++i) {
final double tmp = realEigenvalues[i];
realEigenvalues[i] = realEigenvalues[j];
realEigenvalues[j] = tmp;
--j;
}
}
示例6: include
/**
* The include method is where the QR decomposition occurs. This statement forms all
* intermediate data which will be used for all derivative measures.
* According to the miller paper, note that in the original implementation the x vector
* is overwritten. In this implementation, the include method is passed a copy of the
* original data vector so that there is no contamination of the data. Additionally,
* this method differs slightly from Gentleman's method, in that the assumption is
* of dense design matrices, there is some advantage in using the original gentleman algorithm
* on sparse matrices.
*
* @param x observations on the regressors
* @param wi weight of the this observation (-1,1)
* @param yi observation on the regressand
*/
private void include(final double[] x, final double wi, final double yi) {
int nextr = 0;
double w = wi;
double y = yi;
double xi;
double di;
double wxi;
double dpi;
double xk;
double _w;
this.rss_set = false;
sumy = smartAdd(yi, sumy);
sumsqy = smartAdd(sumsqy, yi * yi);
for (int i = 0; i < x.length; i++) {
if (w == 0.0) {
return;
}
xi = x[i];
if (xi == 0.0) {
nextr += nvars - i - 1;
continue;
}
di = d[i];
wxi = w * xi;
_w = w;
if (di != 0.0) {
dpi = smartAdd(di, wxi * xi);
double tmp = wxi * xi / di;
if (FastMath.abs(tmp) > MathUtils.EPSILON) {
w = (di * w) / dpi;
}
} else {
dpi = wxi * xi;
w = 0.0;
}
d[i] = dpi;
for (int k = i + 1; k < nvars; k++) {
xk = x[k];
x[k] = smartAdd(xk, -xi * r[nextr]);
if (di != 0.0) {
r[nextr] = smartAdd(di * r[nextr], (_w * xi) * xk) / dpi;
} else {
r[nextr] = xk / xi;
}
++nextr;
}
xk = y;
y = smartAdd(xk, -xi * rhs[i]);
if (di != 0.0) {
rhs[i] = smartAdd(di * rhs[i], wxi * xk) / dpi;
} else {
rhs[i] = xk / xi;
}
}
sserr = smartAdd(sserr, w * y * y);
return;
}
示例7: findEigenvalues
/**
* Find the realEigenvalues.
* @exception InvalidMatrixException if a block cannot be diagonalized
*/
private void findEigenvalues()
throws InvalidMatrixException {
// compute splitting points
List<Integer> splitIndices = computeSplits();
// find realEigenvalues in each block
realEigenvalues = new double[main.length];
imagEigenvalues = new double[main.length];
int begin = 0;
for (final int end : splitIndices) {
final int n = end - begin;
switch (n) {
case 1:
// apply dedicated method for dimension 1
process1RowBlock(begin);
break;
case 2:
// apply dedicated method for dimension 2
process2RowsBlock(begin);
break;
case 3:
// apply dedicated method for dimension 3
process3RowsBlock(begin);
break;
default:
// choose an initial shift for LDL<sup>T</sup> decomposition
final double[] range = eigenvaluesRange(begin, n);
final double oneFourth = 0.25 * (3 * range[0] + range[1]);
final int oneFourthCount = countEigenValues(oneFourth, begin, n);
final double threeFourth = 0.25 * (range[0] + 3 * range[1]);
final int threeFourthCount = countEigenValues(threeFourth, begin, n);
final boolean chooseLeft = (oneFourthCount - 1) >= (n - threeFourthCount);
final double lambda = chooseLeft ? range[0] : range[1];
tau = (range[1] - range[0]) * MathUtils.EPSILON * n + 2 * minPivot;
// decompose TλI as LDL<sup>T</sup>
ldlTDecomposition(lambda, begin, n);
// apply general dqd/dqds method
processGeneralBlock(n);
// extract realEigenvalues
if (chooseLeft) {
for (int i = 0; i < n; ++i) {
realEigenvalues[begin + i] = lambda + work[4 * i];
}
} else {
for (int i = 0; i < n; ++i) {
realEigenvalues[begin + i] = lambda - work[4 * i];
}
}
}
begin = end;
}
// sort the realEigenvalues in decreasing order
Arrays.sort(realEigenvalues);
int j = realEigenvalues.length - 1;
for (int i = 0; i < j; ++i) {
final double tmp = realEigenvalues[i];
realEigenvalues[i] = realEigenvalues[j];
realEigenvalues[j] = tmp;
--j;
}
}
示例8: findEigenvalues
/**
* Find the realEigenvalues.
* @exception InvalidMatrixException if a block cannot be diagonalized
*/
private void findEigenvalues()
throws InvalidMatrixException {
// compute splitting points
List<Integer> splitIndices = computeSplits();
// find realEigenvalues in each block
realEigenvalues = new double[main.length];
imagEigenvalues = new double[main.length];
int begin = 0;
for (final int end : splitIndices) {
final int n = end - begin;
switch (n) {
case 1:
// apply dedicated method for dimension 1
process1RowBlock(begin);
break;
case 2:
// apply dedicated method for dimension 2
process2RowsBlock(begin);
break;
case 3:
// apply dedicated method for dimension 3
process3RowsBlock(begin);
break;
default:
// choose an initial shift for LDL<sup>T</sup> decomposition
final double[] range = eigenvaluesRange(begin, n);
final double oneFourth = 0.25 * (3 * range[0] + range[1]);
final int oneFourthCount = countEigenValues(oneFourth, begin, n);
final double threeFourth = 0.25 * (range[0] + 3 * range[1]);
final int threeFourthCount = countEigenValues(threeFourth, begin, n);
final boolean chooseLeft = (oneFourthCount - 1) >= (n - threeFourthCount);
final double lambda = chooseLeft ? range[0] : range[1];
tau = (range[1] - range[0]) * MathUtils.EPSILON * n + 2 * minPivot;
// decompose TλI as LDL<sup>T</sup>
ldlTDecomposition(lambda, begin, n);
// apply general dqd/dqds method
processGeneralBlock(n);
// extract realEigenvalues
if (chooseLeft) {
for (int i = 0; i < n; ++i) {
realEigenvalues[begin + i] = lambda + work[4 * i];
}
} else {
for (int i = 0; i < n; ++i) {
realEigenvalues[begin + i] = lambda - work[4 * i];
}
}
}
begin = end;
}
// sort the realEigenvalues in decreasing order
Arrays.sort(realEigenvalues);
for (int i = 0, j = realEigenvalues.length - 1; i < j; ++i, --j) {
final double tmp = realEigenvalues[i];
realEigenvalues[i] = realEigenvalues[j];
realEigenvalues[j] = tmp;
}
}
示例9: MillerUpdatingRegression
/**
* Primary constructor for the MillerUpdatingRegression
*
* @param numberOfVariables maximum number of potential regressors
* @param includeConstant include a constant automatically
*/
public MillerUpdatingRegression(int numberOfVariables, boolean includeConstant) {
this(numberOfVariables, includeConstant, MathUtils.EPSILON);
}