Examples with KNNHeap de.lmu.ifi.dbs.elki.database.ids.KNNHeap used on opensource projects

Example 11 with KNNHeap

use of de.lmu.ifi.dbs.elki.database.ids.KNNHeap in project elki by elki-project.

the class LinearScanDistanceKNNQuery method getKNNForBulkDBIDs.

@Override
public List<KNNList> getKNNForBulkDBIDs(ArrayDBIDs ids, int k) {
    final int size = ids.size();
    final List<KNNHeap> heaps = new ArrayList<>(size);
    for (int i = 0; i < size; i++) {
        heaps.add(DBIDUtil.newHeap(k));
    }
    linearScanBatchKNN(ids, heaps);
    // Serialize heaps
    List<KNNList> result = new ArrayList<>(size);
    for (KNNHeap heap : heaps) {
        result.add(heap.toKNNList());
    }
    return result;
}

Example 12 with KNNHeap

use of de.lmu.ifi.dbs.elki.database.ids.KNNHeap in project elki by elki-project.

the class LinearScanEuclideanDistanceKNNQuery method getKNNForBulkDBIDs.

@Override
public List<KNNList> getKNNForBulkDBIDs(ArrayDBIDs ids, int k) {
    final Relation<? extends O> relation = getRelation();
    final int size = ids.size();
    final List<KNNHeap> heaps = new ArrayList<>(size);
    List<O> objs = new ArrayList<>(size);
    for (DBIDIter iter = ids.iter(); iter.valid(); iter.advance()) {
        heaps.add(DBIDUtil.newHeap(k));
        objs.add(relation.get(iter));
    }
    linearScanBatchKNN(objs, heaps);
    List<KNNList> result = new ArrayList<>(heaps.size());
    for (KNNHeap heap : heaps) {
        result.add(heap.toKNNListSqrt());
    }
    return result;
}

Example 13 with KNNHeap

use of de.lmu.ifi.dbs.elki.database.ids.KNNHeap in project elki by elki-project.

the class MetricalIndexApproximationMaterializeKNNPreprocessor method preprocess.

@Override
protected void preprocess() {
    final Logging log = getLogger();
    DistanceQuery<O> distanceQuery = relation.getDistanceQuery(distanceFunction);
    MetricalIndexTree<O, N, E> index = getMetricalIndex(relation);
    createStorage();
    MeanVariance pagesize = new MeanVariance();
    MeanVariance ksize = new MeanVariance();
    if (log.isVerbose()) {
        log.verbose("Approximating nearest neighbor lists to database objects");
    }
    List<E> leaves = index.getLeaves();
    FiniteProgress progress = getLogger().isVerbose() ? new FiniteProgress("Processing leaf nodes", leaves.size(), getLogger()) : null;
    for (E leaf : leaves) {
        N node = index.getNode(leaf);
        int size = node.getNumEntries();
        pagesize.put(size);
        if (log.isDebuggingFinest()) {
            log.debugFinest("NumEntires = " + size);
        }
        // Collect the ids in this node.
        ArrayModifiableDBIDs ids = DBIDUtil.newArray(size);
        for (int i = 0; i < size; i++) {
            ids.add(((LeafEntry) node.getEntry(i)).getDBID());
        }
        Object2DoubleOpenHashMap<DBIDPair> cache = new Object2DoubleOpenHashMap<>((size * size * 3) >> 2);
        cache.defaultReturnValue(Double.NaN);
        for (DBIDIter id = ids.iter(); id.valid(); id.advance()) {
            KNNHeap kNN = DBIDUtil.newHeap(k);
            for (DBIDIter id2 = ids.iter(); id2.valid(); id2.advance()) {
                DBIDPair key = DBIDUtil.newPair(id, id2);
                double d = cache.removeDouble(key);
                if (d == d) {
                    // Not NaN
                    // consume the previous result.
                    kNN.insert(d, id2);
                } else {
                    // compute new and store the previous result.
                    d = distanceQuery.distance(id, id2);
                    kNN.insert(d, id2);
                    // put it into the cache, but with the keys reversed
                    key = DBIDUtil.newPair(id2, id);
                    cache.put(key, d);
                }
            }
            ksize.put(kNN.size());
            storage.put(id, kNN.toKNNList());
        }
        if (log.isDebugging() && cache.size() > 0) {
            log.warning("Cache should be empty after each run, but still has " + cache.size() + " elements.");
        }
        log.incrementProcessed(progress);
    }
    log.ensureCompleted(progress);
    if (log.isVerbose()) {
        log.verbose("Average page size = " + pagesize.getMean() + " +- " + pagesize.getSampleStddev());
        log.verbose("On average, " + ksize.getMean() + " +- " + ksize.getSampleStddev() + " neighbors returned.");
    }
}

Example 14 with KNNHeap

use of de.lmu.ifi.dbs.elki.database.ids.KNNHeap in project elki by elki-project.

the class LinearScanDistanceKNNQuery method getKNNForObject.

@Override
public KNNList getKNNForObject(O obj, int k) {
    final DistanceQuery<O> dq = distanceQuery;
    KNNHeap heap = DBIDUtil.newHeap(k);
    double max = Double.POSITIVE_INFINITY;
    for (DBIDIter iter = getRelation().getDBIDs().iter(); iter.valid(); iter.advance()) {
        final double dist = dq.distance(obj, iter);
        if (dist <= max) {
            max = heap.insert(dist, iter);
        }
    }
    return heap.toKNNList();
}

Example 15 with KNNHeap

use of de.lmu.ifi.dbs.elki.database.ids.KNNHeap in project elki by elki-project.

the class LinearScanDistanceKNNQuery method linearScanBatchKNN.

/**
 * Linear batch knn for arbitrary distance functions.
 *
 * @param ids DBIDs to process
 * @param heaps Heaps to store the results in
 */
private void linearScanBatchKNN(ArrayDBIDs ids, List<KNNHeap> heaps) {
    final DistanceQuery<O> dq = distanceQuery;
    // The distance is computed on database IDs
    for (DBIDIter iter = getRelation().getDBIDs().iter(); iter.valid(); iter.advance()) {
        int index = 0;
        for (DBIDIter iter2 = ids.iter(); iter2.valid(); iter2.advance(), index++) {
            KNNHeap heap = heaps.get(index);
            heap.insert(dq.distance(iter2, iter), iter);
        }
    }
}