Package org.grouplens.lenskit.vectors

Examples of org.grouplens.lenskit.vectors.SparseVector

83
84
85
86
87
88
89
90
91
92
93
    }

    @Override
    public Iterable<Neighbor> getCandidateNeighbors(UserHistory<? extends Event> user, LongSet items) {
        final long uid = user.getUserId();
        SparseVector urs = RatingVectorUserHistorySummarizer.makeRatingVector(user);
        final ImmutableSparseVector nratings = normalizer.normalize(user.getUserId(), urs, null)
                                                   .freeze();
        final LongSet candidates = findCandidateNeighbors(uid, nratings, items);
        logger.debug("found {} candidate neighbors for {}", candidates.size(), uid);
        return new Iterable<Neighbor>() {
View Full Code Here
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
        Stopwatch timer = Stopwatch.createStarted();

        for (int i = 0; i < nitems; i++) {
            assert matrix.size() == i;
            final long rowItem = itemDomain.getKey(i);
            final SparseVector vec1 = buildContext.itemVector(rowItem);

            // Take advantage of sparsity if we can
            LongIterator neighbors = iterationStrategy.neighborIterator(buildContext, rowItem, false);
            currentRow.fill(0);

            // Compute similarities and populate the vector
            while (neighbors.hasNext()) {
                final long colItem = neighbors.nextLong();
                final SparseVector vec2 = buildContext.itemVector(colItem);
                assert currentRow.containsKey(colItem);
                currentRow.set(colItem, similarity.similarity(rowItem, vec1, colItem, vec2));
            }

            // Remove the current item (it is not its own neighbor)
View Full Code Here
88
89
90
91
92
93
94
95
96
97
98
99
        SparseVector[] itemRatings = new SparseVector[n];

        for (int i = 0; i < n; i++) {
            final long item = items.getKey(i);
            ScoredIdListBuilder ratings = itemRatingData.get(item);
            SparseVector v = ratings.buildVector();
            assert v.size() == ratings.size();
            itemRatings[i] = v;
            // release some memory
            ratings.clear();
        }
View Full Code Here
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
        // initialize the transposed array to collect item vector data
        Cursor<UserHistory<Event>> users = userEventDAO.streamEventsByUser();
        try {
            for (UserHistory<Event> user : users) {
                long uid = user.getUserId();
                SparseVector summary = userSummarizer.summarize(user);
                MutableSparseVector normed = summary.mutableCopy();
                normalizer.normalize(uid, summary, normed);

                for (VectorEntry rating : normed) {
                    final long item = rating.getKey();
                    // get the item's rating accumulator
                    ScoredIdListBuilder ivect = itemRatings.get(item);
                    if (ivect == null) {
                        ivect = ScoredIds.newListBuilder(100);
                        itemRatings.put(item, ivect);
                    }
                    ivect.add(uid, rating.getValue());
                }

                // get the item's candidate set
                userItems.put(uid, LongUtils.packedSet(summary.keySet()));
            }
        } finally {
            users.close();
        }
    }
View Full Code Here
96
97
98
99
100
101
102
103
104
105
106
    @Test
    public void testItemScorerNoRating() {
        long[] items = {7, 8};
        ItemItemScorer scorer = session.get(ItemItemScorer.class);
        assertThat(scorer, notNullValue());
        SparseVector scores = scorer.score(5, LongArrayList.wrap(items));
        assertThat(scores, notNullValue());
        assertThat(scores.size(), equalTo(1));
        assertThat(scores.get(7), not(notANumber()));
        assertThat(scores.containsKey(8), equalTo(false));
    }
View Full Code Here
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
    @Test
    public void testItemScorerChannels() {
        long[] items = {7, 8};
        ItemItemScorer scorer = session.get(ItemItemScorer.class);
        assertThat(scorer, notNullValue());
        SparseVector scores = scorer.score(5, LongArrayList.wrap(items));
        assertThat(scores, notNullValue());
        assertThat(scores.size(), equalTo(1));
        assertThat(scores.get(7), not(notANumber()));
        assertThat(scores.getChannelVector(ItemItemScorer.NEIGHBORHOOD_SIZE_SYMBOL).
                get(7), closeTo(1.0, 1.0e-5));
        assertThat(scores.containsKey(8), equalTo(false));

        long[] items2 = {7, 8, 9};
        scorer = session.get(ItemItemScorer.class);
        assertThat(scorer, notNullValue());
        scores = scorer.score(2, LongArrayList.wrap(items2));
        assertThat(scores.getChannelVector(ItemItemScorer.NEIGHBORHOOD_SIZE_SYMBOL).
                get(9), closeTo(3.0, 1.0e-5));  // 1, 7, 8
    }
View Full Code Here
42
43
44
45
46
47
48
49
50
51
52
53
54
55
        long[] userIds = {101, 102, 103, 104};
        double[] ratings1 = {4.0, 3.0, 2.5, 2.0};
        double[] ratings2 = {3.0, 2.5, 4.0, 1.0};
        double[] ratings3 = {5.0, 3.5, 0.5, 1.0};
        double[] ratings4 = {4.5, 3.0, 3.5, 1.5};
        SparseVector v1 = MutableSparseVector.wrap(userIds, ratings1);
        SparseVector v2 = MutableSparseVector.wrap(userIds, ratings2);
        SparseVector v3 = MutableSparseVector.wrap(userIds, ratings3);
        SparseVector v4 = MutableSparseVector.wrap(userIds, ratings4);

        SparseVector[] ratings = { v1, v2, v3, v4 };
        ItemItemBuildContext context = new ItemItemBuildContext(items, ratings,
                                                                new Long2ObjectOpenHashMap<LongSortedSet>());
View Full Code Here
64
65
66
67
68
69
70
71
72
73
74
75
        LongKeyDomain items = LongKeyDomain.create(1,2,3,4);

        long[] userIds = {101, 102, 103, 104};
        double[] ratings1 = {4.0, 3.0, 2.5, 2.0};
        double[] ratings4 = {4.5, 3.0, 3.5, 1.5};
        SparseVector v1 = MutableSparseVector.wrap(userIds, ratings1);
        SparseVector v4 = MutableSparseVector.wrap(userIds, ratings4);

        SparseVector[] ratingMap = {
                v1,
                MutableSparseVector.create(),
                MutableSparseVector.create(),
View Full Code Here
68
69
70
71
72
73
74
75
76
77
78
    public void predict(long user, @Nonnull MutableSparseVector scores) {
        itemScorer.score(user, scores);
        if (baselineScorer != null) {
            LongSet unset = scores.unsetKeySet();
            if (!unset.isEmpty()) {
                SparseVector bscores = baselineScorer.score(user, unset);
                scores.set(bscores);
            }
        }
        quantize(scores);
    }
View Full Code Here
52
53
54
55
56
57
58
59
     * {@inheritDoc}
     * <p>Delegates to {@link #score(long, Collection)}.
     */
    @Override
    public double score(long user, long item) {
        SparseVector v = score(user, LongLists.singleton(item));
        return v.get(item, Double.NaN);
    }
View Full Code Here
TOP

Related Classes of org.grouplens.lenskit.vectors.SparseVector

  • it.unimi.dsi.fastutil.doubles.DoubleArrayList
  • it.unimi.dsi.fastutil.doubles.DoubleIterator
  • it.unimi.dsi.fastutil.ints.IntIterator
  • org.grouplens.lenskit.baseline.ConstantItemScorerTest
  • org.grouplens.lenskit.baseline.MeanScorersTest
  • org.grouplens.lenskit.basic.AbstractGlobalItemScorer
  • org.grouplens.lenskit.basic.AbstractItemScorer
  • org.grouplens.lenskit.basic.AbstractRatingPredictor
  • org.grouplens.lenskit.basic.PrecomputedItemScorer
  • org.grouplens.lenskit.basic.RescoringItemRecommender
    • Copyright © 2018 www.massapicom. All rights reserved.
    All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.