Search in sources :

Example 1 with PredictionInput

use of org.kie.kogito.explainability.model.PredictionInput in project kogito-apps by kiegroup.

the class TestUtils method getSymbolicArithmeticModel.

public static PredictionProvider getSymbolicArithmeticModel() {
    return inputs -> supplyAsync(() -> {
        List<PredictionOutput> predictionOutputs = new LinkedList<>();
        final String OPERAND_FEATURE_NAME = "operand";
        for (PredictionInput predictionInput : inputs) {
            List<Feature> features = predictionInput.getFeatures();
            // Find a valid operand feature, if any
            Optional<String> operand = features.stream().filter(f -> OPERAND_FEATURE_NAME.equals(f.getName())).map(f -> f.getValue().asString()).findFirst();
            if (!operand.isPresent()) {
                throw new IllegalArgumentException("No valid operand found in features");
            }
            final String operandValue = operand.get();
            double result = 0;
            // Apply the found operand to the rest of the features
            for (Feature feature : features) {
                if (!OPERAND_FEATURE_NAME.equals(feature.getName())) {
                    switch(operandValue) {
                        case "+":
                            result += feature.getValue().asNumber();
                            break;
                        case "-":
                            result -= feature.getValue().asNumber();
                            break;
                        case "*":
                            result *= feature.getValue().asNumber();
                            break;
                        case "/":
                            result /= feature.getValue().asNumber();
                            break;
                    }
                }
            }
            PredictionOutput predictionOutput = new PredictionOutput(List.of(new Output("result", Type.NUMBER, new Value(result), 1d)));
            predictionOutputs.add(predictionOutput);
        }
        return predictionOutputs;
    });
}
Also used : Arrays(java.util.Arrays) LimeExplainer(org.kie.kogito.explainability.local.lime.LimeExplainer) Feature(org.kie.kogito.explainability.model.Feature) Prediction(org.kie.kogito.explainability.model.Prediction) Random(java.util.Random) Mockito.when(org.mockito.Mockito.when) Value(org.kie.kogito.explainability.model.Value) Type(org.kie.kogito.explainability.model.Type) PredictionProvider(org.kie.kogito.explainability.model.PredictionProvider) ArrayList(java.util.ArrayList) PredictionInput(org.kie.kogito.explainability.model.PredictionInput) List(java.util.List) Pair(org.apache.commons.lang3.tuple.Pair) Output(org.kie.kogito.explainability.model.Output) CompletableFuture.supplyAsync(java.util.concurrent.CompletableFuture.supplyAsync) Optional(java.util.Optional) ValidationUtils(org.kie.kogito.explainability.utils.ValidationUtils) LinkedList(java.util.LinkedList) Assertions.assertDoesNotThrow(org.junit.jupiter.api.Assertions.assertDoesNotThrow) PredictionOutput(org.kie.kogito.explainability.model.PredictionOutput) Mockito.mock(org.mockito.Mockito.mock) PredictionInput(org.kie.kogito.explainability.model.PredictionInput) Feature(org.kie.kogito.explainability.model.Feature) LinkedList(java.util.LinkedList) PredictionOutput(org.kie.kogito.explainability.model.PredictionOutput) Output(org.kie.kogito.explainability.model.Output) PredictionOutput(org.kie.kogito.explainability.model.PredictionOutput) Value(org.kie.kogito.explainability.model.Value)

Example 2 with PredictionInput

use of org.kie.kogito.explainability.model.PredictionInput in project kogito-apps by kiegroup.

the class TestUtils method getNoisySumModel.

public static PredictionProvider getNoisySumModel(Random rn, double noiseMagnitude) {
    return inputs -> supplyAsync(() -> {
        List<PredictionOutput> predictionOutputs = new LinkedList<>();
        for (PredictionInput predictionInput : inputs) {
            List<Feature> features = predictionInput.getFeatures();
            double result = 0;
            for (int i = 0; i < features.size(); i++) {
                result += features.get(i).getValue().asNumber() + ((rn.nextDouble() - .5) * noiseMagnitude);
            }
            PredictionOutput predictionOutput = new PredictionOutput(List.of(new Output("noisy-sum", Type.NUMBER, new Value(result), 1d)));
            predictionOutputs.add(predictionOutput);
        }
        return predictionOutputs;
    });
}
Also used : Arrays(java.util.Arrays) LimeExplainer(org.kie.kogito.explainability.local.lime.LimeExplainer) Feature(org.kie.kogito.explainability.model.Feature) Prediction(org.kie.kogito.explainability.model.Prediction) Random(java.util.Random) Mockito.when(org.mockito.Mockito.when) Value(org.kie.kogito.explainability.model.Value) Type(org.kie.kogito.explainability.model.Type) PredictionProvider(org.kie.kogito.explainability.model.PredictionProvider) ArrayList(java.util.ArrayList) PredictionInput(org.kie.kogito.explainability.model.PredictionInput) List(java.util.List) Pair(org.apache.commons.lang3.tuple.Pair) Output(org.kie.kogito.explainability.model.Output) CompletableFuture.supplyAsync(java.util.concurrent.CompletableFuture.supplyAsync) Optional(java.util.Optional) ValidationUtils(org.kie.kogito.explainability.utils.ValidationUtils) LinkedList(java.util.LinkedList) Assertions.assertDoesNotThrow(org.junit.jupiter.api.Assertions.assertDoesNotThrow) PredictionOutput(org.kie.kogito.explainability.model.PredictionOutput) Mockito.mock(org.mockito.Mockito.mock) PredictionInput(org.kie.kogito.explainability.model.PredictionInput) PredictionOutput(org.kie.kogito.explainability.model.PredictionOutput) Output(org.kie.kogito.explainability.model.Output) PredictionOutput(org.kie.kogito.explainability.model.PredictionOutput) Value(org.kie.kogito.explainability.model.Value) Feature(org.kie.kogito.explainability.model.Feature) LinkedList(java.util.LinkedList)

Example 3 with PredictionInput

use of org.kie.kogito.explainability.model.PredictionInput in project kogito-apps by kiegroup.

the class TestUtils method getSumThresholdModel.

public static PredictionProvider getSumThresholdModel(double center, double epsilon) {
    return inputs -> supplyAsync(() -> {
        List<PredictionOutput> predictionOutputs = new LinkedList<>();
        for (PredictionInput predictionInput : inputs) {
            List<Feature> features = predictionInput.getFeatures();
            double result = 0;
            for (int i = 0; i < features.size(); i++) {
                result += features.get(i).getValue().asNumber();
            }
            final boolean inside = (result >= center - epsilon && result <= center + epsilon);
            PredictionOutput predictionOutput = new PredictionOutput(List.of(new Output("inside", Type.BOOLEAN, new Value(inside), 1.0 - Math.abs(result - center))));
            predictionOutputs.add(predictionOutput);
        }
        return predictionOutputs;
    });
}
Also used : Arrays(java.util.Arrays) LimeExplainer(org.kie.kogito.explainability.local.lime.LimeExplainer) Feature(org.kie.kogito.explainability.model.Feature) Prediction(org.kie.kogito.explainability.model.Prediction) Random(java.util.Random) Mockito.when(org.mockito.Mockito.when) Value(org.kie.kogito.explainability.model.Value) Type(org.kie.kogito.explainability.model.Type) PredictionProvider(org.kie.kogito.explainability.model.PredictionProvider) ArrayList(java.util.ArrayList) PredictionInput(org.kie.kogito.explainability.model.PredictionInput) List(java.util.List) Pair(org.apache.commons.lang3.tuple.Pair) Output(org.kie.kogito.explainability.model.Output) CompletableFuture.supplyAsync(java.util.concurrent.CompletableFuture.supplyAsync) Optional(java.util.Optional) ValidationUtils(org.kie.kogito.explainability.utils.ValidationUtils) LinkedList(java.util.LinkedList) Assertions.assertDoesNotThrow(org.junit.jupiter.api.Assertions.assertDoesNotThrow) PredictionOutput(org.kie.kogito.explainability.model.PredictionOutput) Mockito.mock(org.mockito.Mockito.mock) PredictionInput(org.kie.kogito.explainability.model.PredictionInput) PredictionOutput(org.kie.kogito.explainability.model.PredictionOutput) Output(org.kie.kogito.explainability.model.Output) PredictionOutput(org.kie.kogito.explainability.model.PredictionOutput) Value(org.kie.kogito.explainability.model.Value) Feature(org.kie.kogito.explainability.model.Feature) LinkedList(java.util.LinkedList)

Example 4 with PredictionInput

use of org.kie.kogito.explainability.model.PredictionInput in project kogito-apps by kiegroup.

the class AggregatedLimeExplainerTest method testExplainWithMetadata.

@ParameterizedTest
@ValueSource(ints = { 0, 1, 2, 3, 4 })
void testExplainWithMetadata(int seed) throws ExecutionException, InterruptedException {
    Random random = new Random();
    random.setSeed(seed);
    PredictionProvider sumSkipModel = TestUtils.getSumSkipModel(1);
    PredictionProviderMetadata metadata = new PredictionProviderMetadata() {

        @Override
        public DataDistribution getDataDistribution() {
            return DataUtils.generateRandomDataDistribution(3, 100, random);
        }

        @Override
        public PredictionInput getInputShape() {
            List<Feature> features = new LinkedList<>();
            features.add(FeatureFactory.newNumericalFeature("f0", 0));
            features.add(FeatureFactory.newNumericalFeature("f1", 0));
            features.add(FeatureFactory.newNumericalFeature("f2", 0));
            return new PredictionInput(features);
        }

        @Override
        public PredictionOutput getOutputShape() {
            List<Output> outputs = new LinkedList<>();
            outputs.add(new Output("sum-but1", Type.BOOLEAN, new Value(false), 0d));
            return new PredictionOutput(outputs);
        }
    };
    AggregatedLimeExplainer aggregatedLimeExplainer = new AggregatedLimeExplainer();
    Map<String, Saliency> explain = aggregatedLimeExplainer.explainFromMetadata(sumSkipModel, metadata).get();
    assertNotNull(explain);
    assertEquals(1, explain.size());
    assertTrue(explain.containsKey("sum-but1"));
    Saliency saliency = explain.get("sum-but1");
    assertNotNull(saliency);
    List<String> collect = saliency.getPositiveFeatures(2).stream().map(FeatureImportance::getFeature).map(Feature::getName).collect(Collectors.toList());
    // skipped feature should not appear in top two positive features
    assertFalse(collect.contains("f1"));
}
Also used : PredictionInput(org.kie.kogito.explainability.model.PredictionInput) PredictionProviderMetadata(org.kie.kogito.explainability.model.PredictionProviderMetadata) Saliency(org.kie.kogito.explainability.model.Saliency) PredictionProvider(org.kie.kogito.explainability.model.PredictionProvider) Feature(org.kie.kogito.explainability.model.Feature) LinkedList(java.util.LinkedList) Random(java.util.Random) FeatureImportance(org.kie.kogito.explainability.model.FeatureImportance) PredictionOutput(org.kie.kogito.explainability.model.PredictionOutput) PredictionOutput(org.kie.kogito.explainability.model.PredictionOutput) Output(org.kie.kogito.explainability.model.Output) Value(org.kie.kogito.explainability.model.Value) ValueSource(org.junit.jupiter.params.provider.ValueSource) ParameterizedTest(org.junit.jupiter.params.ParameterizedTest)

Example 5 with PredictionInput

use of org.kie.kogito.explainability.model.PredictionInput in project kogito-apps by kiegroup.

the class CounterfactualEntityFactoryTest method testCreateFixedEntities.

@Test
void testCreateFixedEntities() {
    List<Feature> features = new LinkedList<>();
    features.add(FeatureFactory.newNumericalFeature("f-num1", 100.1));
    features.add(FeatureFactory.newNumericalFeature("f-num2", 100.2, NumericalFeatureDomain.create(0.0, 1000.0)));
    features.add(FeatureFactory.newNumericalFeature("f-num3", 100.3));
    features.add(FeatureFactory.newNumericalFeature("f-num4", 100.4, NumericalFeatureDomain.create(0.0, 1000.0)));
    PredictionInput input = new PredictionInput(features);
    List<CounterfactualEntity> entities = CounterfactualEntityFactory.createEntities(input);
    // Check types
    assertTrue(entities.get(0) instanceof FixedDoubleEntity);
    assertTrue(entities.get(1) instanceof DoubleEntity);
    assertTrue(entities.get(2) instanceof FixedDoubleEntity);
    assertTrue(entities.get(3) instanceof DoubleEntity);
    // Check values
    assertEquals(100.1, entities.get(0).asFeature().getValue().asNumber());
    assertEquals(100.2, entities.get(1).asFeature().getValue().asNumber());
    assertEquals(100.3, entities.get(2).asFeature().getValue().asNumber());
    assertEquals(100.4, entities.get(3).asFeature().getValue().asNumber());
    // Check constraints
    assertTrue(entities.get(0).isConstrained());
    assertFalse(entities.get(1).isConstrained());
    assertTrue(entities.get(2).isConstrained());
    assertFalse(entities.get(3).isConstrained());
}
Also used : CounterfactualEntity(org.kie.kogito.explainability.local.counterfactual.entities.CounterfactualEntity) PredictionInput(org.kie.kogito.explainability.model.PredictionInput) FixedDoubleEntity(org.kie.kogito.explainability.local.counterfactual.entities.fixed.FixedDoubleEntity) DoubleEntity(org.kie.kogito.explainability.local.counterfactual.entities.DoubleEntity) FixedDoubleEntity(org.kie.kogito.explainability.local.counterfactual.entities.fixed.FixedDoubleEntity) Feature(org.kie.kogito.explainability.model.Feature) LinkedList(java.util.LinkedList) Test(org.junit.jupiter.api.Test)

Aggregations

PredictionInput (org.kie.kogito.explainability.model.PredictionInput)187 PredictionOutput (org.kie.kogito.explainability.model.PredictionOutput)143 PredictionProvider (org.kie.kogito.explainability.model.PredictionProvider)135 Prediction (org.kie.kogito.explainability.model.Prediction)126 Feature (org.kie.kogito.explainability.model.Feature)109 Test (org.junit.jupiter.api.Test)107 ArrayList (java.util.ArrayList)97 SimplePrediction (org.kie.kogito.explainability.model.SimplePrediction)95 Random (java.util.Random)86 PerturbationContext (org.kie.kogito.explainability.model.PerturbationContext)67 ParameterizedTest (org.junit.jupiter.params.ParameterizedTest)60 Output (org.kie.kogito.explainability.model.Output)55 LimeConfig (org.kie.kogito.explainability.local.lime.LimeConfig)54 LinkedList (java.util.LinkedList)53 LimeExplainer (org.kie.kogito.explainability.local.lime.LimeExplainer)52 Value (org.kie.kogito.explainability.model.Value)52 Saliency (org.kie.kogito.explainability.model.Saliency)50 List (java.util.List)39 LimeConfigOptimizer (org.kie.kogito.explainability.local.lime.optim.LimeConfigOptimizer)33 Type (org.kie.kogito.explainability.model.Type)31