public MapVector sparsify(MapVector input, final int numNonZeroEntries) {
Comparator<IntDoublePair> comparator =
new Comparator<IntDoublePair>() {
public int compare(IntDoublePair o1, IntDoublePair o2) {
return Math.abs(o1.getDouble()) - Math.abs(o2.getDouble()) > 0 ? 1 : -1;
}
};
Set<IntDoublePair> bestPairs =
new FixedSizeSortedSet<IntDoublePair>(comparator, numNonZeroEntries);
for (final IntDoublePair pair : input) bestPairs.add(new IntDoublePairImpl(pair));
List<IntDoublePair> resortedPairs = new ArrayList<IntDoublePair>(bestPairs);
Collections.sort(
resortedPairs,
new Comparator<IntDoublePair>() {
public int compare(IntDoublePair o1, IntDoublePair o2) {
if (o1.getInt() == o2.getInt()) return 0;
return o1.getInt() - o2.getInt() > 0 ? 1 : -1;
}
});
int[] indices = new int[numNonZeroEntries];
double[] values = new double[numNonZeroEntries];
int i = 0;
for (IntDoublePair pair : resortedPairs) {
indices[i] = pair.getInt();
values[i] = pair.getDouble();
i++;
if (i >= numNonZeroEntries) break;
}
return new ImmutableSparseMapVector(indices, values);
}
public MapVector project(MapVector documentVector) {
MapVector output = _eigenvectors.times(documentVector);
for (IntDoublePair pair : output) {
if (pair.getInt() >= _inverseSingularValues.length) break;
pair.setDouble(pair.getDouble() * _inverseSingularValues[pair.getInt()]);
}
return output;
}