private static DenseMatrix64F multByExtract(
int operationType, D1Submatrix64F subA, D1Submatrix64F subB, D1Submatrix64F subC) {
SimpleMatrix A = subA.extract();
SimpleMatrix B = subB.extract();
SimpleMatrix C = subC.extract();
if (operationType > 0) return A.mult(B).plus(C).getMatrix();
else if (operationType < 0) return C.minus(A.mult(B)).getMatrix();
else return A.mult(B).getMatrix();
}
public int predictState(int from_state, int number_of_steps) {
long lStartTime = System.currentTimeMillis();
int from_bin = getBinNumber((float) from_state);
SimpleMatrix result = new SimpleMatrix(transition_matrix);
for (int i = 1; i < number_of_steps; i++) result = result.mult(transition_matrix);
SimpleMatrix row_result = result.extractVector(true, from_bin);
// row_result.print();
double maxVal = row_result.get(0, 0);
int maxValIndex = 0;
for (int i = 1; i < row_result.numCols(); i++) {
if (row_result.get(0, i) >= maxVal) {
maxVal = row_result.get(0, i);
maxValIndex = i;
}
}
long lEndTime = System.currentTimeMillis();
logger.info(
("From: "
+ from_state
+ " In NumberOfSteps: "
+ number_of_steps
+ " MaxProbabilityBin: "
+ maxValIndex
+ " MostProbablyTo: "
+ getOrgValFromBinNumber(maxValIndex)));
logger.info(("Time taken for prediction = " + (lEndTime - lStartTime)));
return getOrgValFromBinNumber(maxValIndex);
}
private void checkDecomposition(int height, int width, boolean compact) {
QRDecomposition<DenseMatrix64F> alg = createQRDecomposition();
SimpleMatrix A = new SimpleMatrix(height, width);
RandomMatrices.setRandom(A.getMatrix(), rand);
assertTrue(alg.decompose(A.copy().getMatrix()));
int minStride = Math.min(height, width);
SimpleMatrix Q = new SimpleMatrix(height, compact ? minStride : height);
alg.getQ(Q.getMatrix(), compact);
SimpleMatrix R = new SimpleMatrix(compact ? minStride : height, width);
alg.getR(R.getMatrix(), compact);
// see if Q has the expected properties
assertTrue(MatrixFeatures.isOrthogonal(Q.getMatrix(), 1e-6));
// UtilEjml.print(alg.getQR());
// Q.print();
// R.print();
// see if it has the expected properties
DenseMatrix64F A_found = Q.mult(R).getMatrix();
EjmlUnitTests.assertEquals(A.getMatrix(), A_found, 1e-6);
assertTrue(Q.transpose().mult(A).isIdentical(R, 1e-6));
}
@Override
public double[][] predict(List<PredictionPaper> testDocs) {
String testData = "lda/test.dat";
createLdaInputTest(testData, testDocs);
Utils.runCommand(
"lib/lda-c-dist/lda inf "
+ " lib/lda-c-dist/settings.txt "
+ "lda/final "
+ testData
+ " lda/output",
false);
double[][] gammasMatrix = Utils.readMatrix("lda/output-gamma.dat", false);
double alpha = Utils.readAlpha("lda/final.other");
for (int i = 0; i < gammasMatrix.length; i++) {
for (int j = 0; j < gammasMatrix[i].length; j++) {
gammasMatrix[i][j] -= alpha;
}
}
SimpleMatrix gammas = new SimpleMatrix(gammasMatrix);
SimpleMatrix beta = new SimpleMatrix(betaMatrix);
SimpleMatrix probabilities = gammas.mult(beta);
double[][] result = new double[probabilities.numRows()][probabilities.numCols()];
for (int row = 0; row < probabilities.numRows(); row++) {
for (int col = 0; col < probabilities.numCols(); col++) {
result[row][col] = probabilities.get(row, col);
}
}
return result;
}
public void translate(double x, double y) {
matrix =
matrix.mult(
new SimpleMatrix(
new double[][] {
{1, 0, 0},
{0, 1, 0},
{x, y, 1}
}));
}
public void rotate(Vector heading, Vector side) {
matrix =
matrix.mult(
new SimpleMatrix(
new double[][] {
{heading.X(), heading.Y(), 0},
{side.X(), side.Y(), 0},
{0, 0, 1}
}));
}
/** H0 = H*M P=[M|m] from canonical camera */
private SimpleMatrix computeHZero(DenseMatrix64F F, Point3D_F64 e2, SimpleMatrix H) {
Vector3D_F64 v = new Vector3D_F64(.1, 0.5, .2);
// need to make sure M is not singular for this technique to work
SimpleMatrix P = SimpleMatrix.wrap(MultiViewOps.canonicalCamera(F, e2, v, 1));
SimpleMatrix M = P.extractMatrix(0, 3, 0, 3);
return H.mult(M);
}
public void rotate(double angle) {
double sin = Math.sin(angle);
double cos = Math.cos(angle);
matrix =
matrix.mult(
new SimpleMatrix(
new double[][] {
{cos, sin, 0},
{-sin, cos, 0},
{0, 0, 1}
}));
}
/**
* Compute rectification transforms for the stereo pair given a fundamental matrix and its
* observations.
*
* @param F Fundamental matrix
* @param observations Observations used to compute F
* @param width Width of first image.
* @param height Height of first image.
*/
public void process(DenseMatrix64F F, List<AssociatedPair> observations, int width, int height) {
int centerX = width / 2;
int centerY = height / 2;
MultiViewOps.extractEpipoles(F, epipole1, epipole2);
checkEpipoleInside(width, height);
// compute the transform H which will send epipole2 to infinity
SimpleMatrix R = rotateEpipole(epipole2, centerX, centerY);
SimpleMatrix T = translateToOrigin(centerX, centerY);
SimpleMatrix G = computeG(epipole2, centerX, centerY);
SimpleMatrix H = G.mult(R).mult(T);
// Find the two matching transforms
SimpleMatrix Hzero = computeHZero(F, epipole2, H);
SimpleMatrix Ha = computeAffineH(observations, H.getMatrix(), Hzero.getMatrix());
rect1.set(Ha.mult(Hzero).getMatrix());
rect2.set(H.getMatrix());
}
public void backpropDerivative(
Tree tree,
List<String> words,
IdentityHashMap<Tree, SimpleMatrix> nodeVectors,
TwoDimensionalMap<String, String, SimpleMatrix> binaryW_dfs,
Map<String, SimpleMatrix> unaryW_dfs,
TwoDimensionalMap<String, String, SimpleMatrix> binaryScoreDerivatives,
Map<String, SimpleMatrix> unaryScoreDerivatives,
Map<String, SimpleMatrix> wordVectorDerivatives,
SimpleMatrix deltaUp) {
if (tree.isLeaf()) {
return;
}
if (tree.isPreTerminal()) {
if (op.trainOptions.trainWordVectors) {
String word = tree.children()[0].label().value();
word = dvModel.getVocabWord(word);
// SimpleMatrix currentVector = nodeVectors.get(tree);
// SimpleMatrix currentVectorDerivative =
// nonlinearityVectorToDerivative(currentVector);
// SimpleMatrix derivative = deltaUp.elementMult(currentVectorDerivative);
SimpleMatrix derivative = deltaUp;
wordVectorDerivatives.put(word, wordVectorDerivatives.get(word).plus(derivative));
}
return;
}
SimpleMatrix currentVector = nodeVectors.get(tree);
SimpleMatrix currentVectorDerivative =
NeuralUtils.elementwiseApplyTanhDerivative(currentVector);
SimpleMatrix scoreW = dvModel.getScoreWForNode(tree);
currentVectorDerivative = currentVectorDerivative.elementMult(scoreW.transpose());
// the delta that is used at the current nodes
SimpleMatrix deltaCurrent = deltaUp.plus(currentVectorDerivative);
SimpleMatrix W = dvModel.getWForNode(tree);
SimpleMatrix WTdelta = W.transpose().mult(deltaCurrent);
if (tree.children().length == 2) {
// TODO: RS: Change to the nice "getWForNode" setup?
String leftLabel = dvModel.basicCategory(tree.children()[0].label().value());
String rightLabel = dvModel.basicCategory(tree.children()[1].label().value());
binaryScoreDerivatives.put(
leftLabel,
rightLabel,
binaryScoreDerivatives.get(leftLabel, rightLabel).plus(currentVector.transpose()));
SimpleMatrix leftVector = nodeVectors.get(tree.children()[0]);
SimpleMatrix rightVector = nodeVectors.get(tree.children()[1]);
SimpleMatrix childrenVector = NeuralUtils.concatenateWithBias(leftVector, rightVector);
if (op.trainOptions.useContextWords) {
childrenVector = concatenateContextWords(childrenVector, tree.getSpan(), words);
}
SimpleMatrix W_df = deltaCurrent.mult(childrenVector.transpose());
binaryW_dfs.put(leftLabel, rightLabel, binaryW_dfs.get(leftLabel, rightLabel).plus(W_df));
// and then recurse
SimpleMatrix leftDerivative = NeuralUtils.elementwiseApplyTanhDerivative(leftVector);
SimpleMatrix rightDerivative = NeuralUtils.elementwiseApplyTanhDerivative(rightVector);
SimpleMatrix leftWTDelta = WTdelta.extractMatrix(0, deltaCurrent.numRows(), 0, 1);
SimpleMatrix rightWTDelta =
WTdelta.extractMatrix(deltaCurrent.numRows(), deltaCurrent.numRows() * 2, 0, 1);
backpropDerivative(
tree.children()[0],
words,
nodeVectors,
binaryW_dfs,
unaryW_dfs,
binaryScoreDerivatives,
unaryScoreDerivatives,
wordVectorDerivatives,
leftDerivative.elementMult(leftWTDelta));
backpropDerivative(
tree.children()[1],
words,
nodeVectors,
binaryW_dfs,
unaryW_dfs,
binaryScoreDerivatives,
unaryScoreDerivatives,
wordVectorDerivatives,
rightDerivative.elementMult(rightWTDelta));
} else if (tree.children().length == 1) {
String childLabel = dvModel.basicCategory(tree.children()[0].label().value());
unaryScoreDerivatives.put(
childLabel, unaryScoreDerivatives.get(childLabel).plus(currentVector.transpose()));
SimpleMatrix childVector = nodeVectors.get(tree.children()[0]);
SimpleMatrix childVectorWithBias = NeuralUtils.concatenateWithBias(childVector);
if (op.trainOptions.useContextWords) {
childVectorWithBias = concatenateContextWords(childVectorWithBias, tree.getSpan(), words);
}
SimpleMatrix W_df = deltaCurrent.mult(childVectorWithBias.transpose());
// System.out.println("unary backprop derivative for " + childLabel);
// System.out.println("Old transform:");
// System.out.println(unaryW_dfs.get(childLabel));
// System.out.println(" Delta:");
// System.out.println(W_df.scale(scale));
unaryW_dfs.put(childLabel, unaryW_dfs.get(childLabel).plus(W_df));
// and then recurse
SimpleMatrix childDerivative = NeuralUtils.elementwiseApplyTanhDerivative(childVector);
// SimpleMatrix childDerivative = childVector;
SimpleMatrix childWTDelta = WTdelta.extractMatrix(0, deltaCurrent.numRows(), 0, 1);
backpropDerivative(
tree.children()[0],
words,
nodeVectors,
binaryW_dfs,
unaryW_dfs,
binaryScoreDerivatives,
unaryScoreDerivatives,
wordVectorDerivatives,
childDerivative.elementMult(childWTDelta));
}
}
private void forwardPropagateTree(
Tree tree,
List<String> words,
IdentityHashMap<Tree, SimpleMatrix> nodeVectors,
IdentityHashMap<Tree, Double> scores) {
if (tree.isLeaf()) {
return;
}
if (tree.isPreTerminal()) {
Tree wordNode = tree.children()[0];
String word = wordNode.label().value();
SimpleMatrix wordVector = dvModel.getWordVector(word);
wordVector = NeuralUtils.elementwiseApplyTanh(wordVector);
nodeVectors.put(tree, wordVector);
return;
}
for (Tree child : tree.children()) {
forwardPropagateTree(child, words, nodeVectors, scores);
}
// at this point, nodeVectors contains the vectors for all of
// the children of tree
SimpleMatrix childVec;
if (tree.children().length == 2) {
childVec =
NeuralUtils.concatenateWithBias(
nodeVectors.get(tree.children()[0]), nodeVectors.get(tree.children()[1]));
} else {
childVec = NeuralUtils.concatenateWithBias(nodeVectors.get(tree.children()[0]));
}
if (op.trainOptions.useContextWords) {
childVec = concatenateContextWords(childVec, tree.getSpan(), words);
}
SimpleMatrix W = dvModel.getWForNode(tree);
if (W == null) {
String error = "Could not find W for tree " + tree;
if (op.testOptions.verbose) {
System.err.println(error);
}
throw new NoSuchParseException(error);
}
SimpleMatrix currentVector = W.mult(childVec);
currentVector = NeuralUtils.elementwiseApplyTanh(currentVector);
nodeVectors.put(tree, currentVector);
SimpleMatrix scoreW = dvModel.getScoreWForNode(tree);
if (scoreW == null) {
String error = "Could not find scoreW for tree " + tree;
if (op.testOptions.verbose) {
System.err.println(error);
}
throw new NoSuchParseException(error);
}
double score = scoreW.dot(currentVector);
// score = NeuralUtils.sigmoid(score);
scores.put(tree, score);
// System.err.print(Double.toString(score)+" ");
}
