Examples with LabeledVector org.apache.ignite.ml.structures.LabeledVector used on opensource projects

Example 1 with LabeledVector

use of org.apache.ignite.ml.structures.LabeledVector in project ignite by apache.

the class Deltas method calculateUpdates.

/**
 */
private Vector calculateUpdates(Vector weights, Dataset<SVMPartitionContext, LabeledDataset<Double, LabeledVector>> dataset) {
    return dataset.compute(data -> {
        Vector copiedWeights = weights.copy();
        Vector deltaWeights = initializeWeightsWithZeros(weights.size());
        final int amountOfObservation = data.rowSize();
        Vector tmpAlphas = initializeWeightsWithZeros(amountOfObservation);
        Vector deltaAlphas = initializeWeightsWithZeros(amountOfObservation);
        for (int i = 0; i < this.getAmountOfLocIterations(); i++) {
            int randomIdx = ThreadLocalRandom.current().nextInt(amountOfObservation);
            Deltas deltas = getDeltas(data, copiedWeights, amountOfObservation, tmpAlphas, randomIdx);
            // creates new vector
            copiedWeights = copiedWeights.plus(deltas.deltaWeights);
            // creates new vector
            deltaWeights = deltaWeights.plus(deltas.deltaWeights);
            tmpAlphas.set(randomIdx, tmpAlphas.get(randomIdx) + deltas.deltaAlpha);
            deltaAlphas.set(randomIdx, deltaAlphas.get(randomIdx) + deltas.deltaAlpha);
        }
        return deltaWeights;
    }, (a, b) -> a == null ? b : a.plus(b));
}

Example 2 with LabeledVector

use of org.apache.ignite.ml.structures.LabeledVector in project ignite by apache.

the class Deltas method makeVectorWithInterceptElement.

/**
 */
private Vector makeVectorWithInterceptElement(LabeledVector row) {
    Vector vec = row.features().like(row.features().size() + 1);
    // set intercept element
    vec.set(0, 1);
    for (int j = 0; j < row.features().size(); j++) vec.set(j + 1, row.features().get(j));
    return vec;
}

Example 3 with LabeledVector

use of org.apache.ignite.ml.structures.LabeledVector in project ignite by apache.

the class Deltas method getDeltas.

/**
 */
private Deltas getDeltas(LabeledDataset data, Vector copiedWeights, int amountOfObservation, Vector tmpAlphas, int randomIdx) {
    LabeledVector row = (LabeledVector) data.getRow(randomIdx);
    Double lb = (Double) row.label();
    Vector v = makeVectorWithInterceptElement(row);
    double alpha = tmpAlphas.get(randomIdx);
    return maximize(lb, v, alpha, copiedWeights, amountOfObservation);
}

Example 4 with LabeledVector

use of org.apache.ignite.ml.structures.LabeledVector in project ignite by apache.

the class KNNModel method getDistances.

/**
 * Computes distances between given vector and each vector in training dataset.
 *
 * @param v The given vector.
 * @param trainingData The training dataset.
 * @return Key - distanceMeasure from given features before features with idx stored in value. Value is presented
 * with Set because there can be a few vectors with the same distance.
 */
@NotNull
private TreeMap<Double, Set<Integer>> getDistances(Vector v, LabeledVector[] trainingData) {
    TreeMap<Double, Set<Integer>> distanceIdxPairs = new TreeMap<>();
    for (int i = 0; i < trainingData.length; i++) {
        LabeledVector labeledVector = trainingData[i];
        if (labeledVector != null) {
            double distance = distanceMeasure.compute(v, labeledVector.features());
            putDistanceIdxPair(distanceIdxPairs, i, distance);
        }
    }
    return distanceIdxPairs;
}

Example 5 with LabeledVector

use of org.apache.ignite.ml.structures.LabeledVector in project ignite by apache.

the class MLPGroupTrainerTest method doTestXOR.

/**
 * Test training of 'xor' by {@link MLPGroupUpdateTrainer}.
 */
private <U extends Serializable> void doTestXOR(UpdatesStrategy<? super MultilayerPerceptron, U> stgy) {
    int samplesCnt = 1000;
    Matrix xorInputs = new DenseLocalOnHeapMatrix(new double[][] { { 0.0, 0.0 }, { 0.0, 1.0 }, { 1.0, 0.0 }, { 1.0, 1.0 } }, StorageConstants.ROW_STORAGE_MODE).transpose();
    Matrix xorOutputs = new DenseLocalOnHeapMatrix(new double[][] { { 0.0 }, { 1.0 }, { 1.0 }, { 0.0 } }, StorageConstants.ROW_STORAGE_MODE).transpose();
    MLPArchitecture conf = new MLPArchitecture(2).withAddedLayer(10, true, Activators.RELU).withAddedLayer(1, false, Activators.SIGMOID);
    IgniteCache<Integer, LabeledVector<Vector, Vector>> cache = LabeledVectorsCache.createNew(ignite);
    String cacheName = cache.getName();
    Random rnd = new Random(12345L);
    try (IgniteDataStreamer<Integer, LabeledVector<Vector, Vector>> streamer = ignite.dataStreamer(cacheName)) {
        streamer.perNodeBufferSize(10000);
        for (int i = 0; i < samplesCnt; i++) {
            int col = Math.abs(rnd.nextInt()) % 4;
            streamer.addData(i, new LabeledVector<>(xorInputs.getCol(col), xorOutputs.getCol(col)));
        }
    }
    int totalCnt = 30;
    int failCnt = 0;
    double maxFailRatio = 0.3;
    MLPGroupUpdateTrainer<U> trainer = MLPGroupUpdateTrainer.getDefault(ignite).withSyncPeriod(3).withTolerance(0.001).withMaxGlobalSteps(100).withUpdateStrategy(stgy);
    for (int i = 0; i < totalCnt; i++) {
        MLPGroupUpdateTrainerCacheInput trainerInput = new MLPGroupUpdateTrainerCacheInput(conf, new RandomInitializer(new Random(123L + i)), 6, cache, 10, new Random(123L + i));
        MultilayerPerceptron mlp = trainer.train(trainerInput);
        Matrix predict = mlp.apply(xorInputs);
        Tracer.showAscii(predict);
        X.println(xorOutputs.getRow(0).minus(predict.getRow(0)).kNorm(2) + "");
        failCnt += TestUtils.checkIsInEpsilonNeighbourhoodBoolean(xorOutputs.getRow(0), predict.getRow(0), 5E-1) ? 0 : 1;
    }
    double failRatio = (double) failCnt / totalCnt;
    System.out.println("Fail percentage: " + (failRatio * 100) + "%.");
    assertTrue(failRatio < maxFailRatio);
}