Package com.etsy.conjecture.data

Examples of com.etsy.conjecture.data.StringKeyedVector

42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
        }
    }

    @Override
    public StringKeyedVector getGradients(LabeledInstance<BinaryLabel> instance) {
        StringKeyedVector gradients = instance.getVector().copy();
        double inner = param.dot(instance.getVector());
        double label = instance.getLabel().getAsPlusMinus();
        double z = inner * label;
        if (z <= this.threshold) {
            gradients.mul(-label);
            return gradients;
        } else {
            return new StringKeyedVector();
        }       
    }
View Full Code Here
22
23
24
25
26
27
28
29
30
  private static final long serialVersionUID = 1L;
  private Map<String, Double> clusterCounts = new HashMap<String, Double>();

  public KMeans(String[] categories) {
    for(String s : categories) {
        param.put(s, new StringKeyedVector());
        clusterCounts.put(s, 0.0);
    }
  }
View Full Code Here
54
55
56
57
58
59
60
61
62
63
64
65
    Double current_count = clusterCounts.get(closest_center);
    clusterCounts.put(closest_center, current_count+1.0);
    // Get per center learning rate
    Double learning_rate = 1.0/clusterCounts.get(closest_center);
    // take gradient step
    StringKeyedVector center = param.get(closest_center);
    center.mul(1-learning_rate);
    instance.mul(learning_rate);
    center.add(instance);
    l1Projection(center);
    param.put(closest_center, center);
  }
View Full Code Here
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
    private StringKeyedVector unnormalizedGradients = new StringKeyedVector();
    private StringKeyedVector summedGradients = new StringKeyedVector();

    @Override
    public StringKeyedVector getUpdate(LabeledInstance instance) {
        StringKeyedVector gradients = model.getGradients(instance);
        StringKeyedVector updateVec = new StringKeyedVector();
        Iterator<Map.Entry<String, Double>> it = gradients.iterator();
        while (it.hasNext()) {
            Map.Entry<String,Double> pairs = (Map.Entry)it.next();
            String feature = pairs.getKey();
            double gradient = pairs.getValue();
            double featureLearningRate = updateAndGetFeatureLearningRate(feature, gradient);
            updateVec.setCoordinate(feature, gradient * -featureLearningRate);
       }
       return updateVec;
    }
View Full Code Here
92
93
94
95
96
97
98
99
        return param;
    }

    @Override
    public void teardown() {
        summedGradients = new StringKeyedVector();
        unnormalizedGradients = new StringKeyedVector();
    }
View Full Code Here
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
        return new BinaryLabel(Utilities.logistic(inner));
    }

    @Override
    public StringKeyedVector getGradients(LabeledInstance<BinaryLabel> instance) {
        StringKeyedVector gradients = instance.getVector().copy();
        double label = instance.getLabel().getAsPlusMinus();
        double prediction = param.dot(instance.getVector());
        double loss = Math.max(0, 1d - label * prediction);
        if (loss > 0) {
            double norm = instance.getVector().LPNorm(2d);
            double tau = loss / (norm * norm);
            gradients.mul(tau * label);
            return gradients;
        } else {
          return new StringKeyedVector();
        }
    }
View Full Code Here
9
10
11
12
13
14
15
16
17
18
public class ElasticNetOptimizer extends SGDOptimizer implements LazyVector.UpdateFunction {

    @Override
    public StringKeyedVector getUpdate(LabeledInstance instance) {
        StringKeyedVector gradients = model.getGradients(instance);
        double learningRate = getDecreasingLearningRate(model.epoch);
        gradients.mul(-learningRate);
        return gradients;
    }
View Full Code Here
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
    private StringKeyedVector n = new StringKeyedVector();

    @Override
    public StringKeyedVector getUpdate(LabeledInstance<L> instance) {
        FTRLRegularization(instance);
        StringKeyedVector gradients = model.getGradients(instance);
        Iterator<Map.Entry<String, Double>> it = gradients.iterator();
        while (it.hasNext()) {
            Map.Entry<String,Double> pairs = (Map.Entry)it.next();
            String feature = pairs.getKey();
            double gradient = pairs.getValue();
            double eta = getFeatureLearningRate(feature, gradient);
            double z_i = 0.0; // if first round, set z_i to 0.0
            if (z.containsKey(feature)) {
                z_i = z.getCoordinate(feature);
            }
            double update = (z_i + gradient) - eta * model.param.getCoordinate(feature);
            z.setCoordinate(feature, update);
            double n_i = 0.0; // if first round, set n_i to 0.0
            if (n.containsKey(feature)) {
                n_i = n.getCoordinate(feature);
            }
            n.setCoordinate(feature, n_i + gradient * gradient);
       }
       return new StringKeyedVector(); // Model updates happen in the FTRLRegularization step
    }
View Full Code Here
105
106
107
108
109
110
111
112
        return this;
    }

    @Override
    public void teardown() {
        z = new StringKeyedVector();
        n = new StringKeyedVector();
    }
View Full Code Here
93
94
95
96
97
98
99
100
101
102
        assertTrue(p.predict(getNegativeInstance().getVector()).getValue() < 0.5);
    }

    public void testInstanceNotModified(UpdateableLinearModel model) {
        BinaryLabeledInstance instance = getPositiveInstance();
        StringKeyedVector instanceCopy = instance.getVector().copy();
        model.update(instance);
        assertEquals(instance.getVector().getCoordinate("foo"), instanceCopy.getCoordinate("foo"), 0.0);
        assertEquals(instance.getVector().getCoordinate("bar"), instanceCopy.getCoordinate("bar"), 0.0);
    }
View Full Code Here
TOP

Related Classes of com.etsy.conjecture.data.StringKeyedVector

  • com.etsy.conjecture.model.AdagradOptimizer
  • com.etsy.conjecture.model.ClusteringModel
  • com.etsy.conjecture.model.ControlOptimizer
  • com.etsy.conjecture.model.ElasticNetOptimizer
  • com.etsy.conjecture.model.FTRLOptimizer
  • com.etsy.conjecture.model.Hinge
  • com.etsy.conjecture.model.KMeans
  • com.etsy.conjecture.model.LeastSquaresRegressionModel
  • com.etsy.conjecture.model.LogisticRegression
  • com.etsy.conjecture.model.MIRA
    • 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.