Examples with PerturbationContext org.kie.kogito.explainability.model.PerturbationContext used on opensource projects

Example 56 with PerturbationContext

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

the class LimeConfigEntityFactory method initProcessors.

private static Map<String, BiFunction<LimeConfig, LimeConfigEntity, LimeConfig>> initProcessors() {
    Map<String, BiFunction<LimeConfig, LimeConfigEntity, LimeConfig>> processors = new HashMap<>();
    processors.put(PROXIMITY_KERNEL_WIDTH, (limeConfig, limeConfigEntity) -> limeConfig.withProximityKernelWidth(limeConfigEntity.asDouble()));
    processors.put(PROXIMITY_THRESHOLD, (limeConfig, limeConfigEntity) -> limeConfig.withProximityThreshold(limeConfigEntity.asDouble()));
    processors.put(PROXIMITY_FILTERED_DATASET_MINIMUM, (limeConfig, limeConfigEntity) -> limeConfig.withProximityFilteredDatasetMinimum(limeConfigEntity.asDouble()));
    processors.put(EP_NUMERIC_CLUSTER_FILTER_WIDTH, (limeConfig, limeConfigEntity) -> limeConfig.withEncodingParams(new EncodingParams(limeConfigEntity.asDouble(), limeConfig.getEncodingParams().getNumericTypeClusterThreshold())));
    processors.put(EP_NUMERIC_CLUSTER_THRESHOLD, (limeConfig, limeConfigEntity) -> limeConfig.withEncodingParams(new EncodingParams(limeConfig.getEncodingParams().getNumericTypeClusterGaussianFilterWidth(), limeConfigEntity.asDouble())));
    processors.put(SAMPLING_SEPARABLE_DATASET_RATIO, (limeConfig, limeConfigEntity) -> limeConfig.withSeparableDatasetRatio(limeConfigEntity.asDouble()));
    processors.put(SAMPLING_SIZE, (limeConfig, limeConfigEntity) -> limeConfig.withSamples((int) limeConfigEntity.asDouble()));
    processors.put(SAMPLING_PERTURBATIONS, (limeConfig, limeConfigEntity) -> limeConfig.withPerturbationContext(limeConfig.getPerturbationContext().getSeed().isPresent() ? new PerturbationContext(limeConfig.getPerturbationContext().getSeed().get(), limeConfig.getPerturbationContext().getRandom(), (int) limeConfigEntity.asDouble()) : new PerturbationContext(limeConfig.getPerturbationContext().getRandom(), (int) limeConfigEntity.asDouble())));
    processors.put(PROXIMITY_FILTER_ENABLED, (limeConfig, limeConfigEntity) -> limeConfig.withProximityFilter(limeConfigEntity.asBoolean()));
    processors.put(WEIGHTING_PENALIZE_BALANCE_SPARSE, (limeConfig, limeConfigEntity) -> limeConfig.withPenalizeBalanceSparse(limeConfigEntity.asBoolean()));
    processors.put(SAMPLING_ADAPT_DATASET_VARIANCE, (limeConfig, limeConfigEntity) -> limeConfig.withAdaptiveVariance(limeConfigEntity.asBoolean()));
    return processors;
}

Example 57 with PerturbationContext

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

the class TrafficViolationDmnLimeExplainerTest method testTrafficViolationDMNExplanation.

@Test
void testTrafficViolationDMNExplanation() throws ExecutionException, InterruptedException, TimeoutException {
    PredictionProvider model = getModel();
    PredictionInput predictionInput = getTestInput();
    List<PredictionOutput> predictionOutputs = model.predictAsync(List.of(predictionInput)).get(Config.INSTANCE.getAsyncTimeout(), Config.INSTANCE.getAsyncTimeUnit());
    Prediction prediction = new SimplePrediction(predictionInput, predictionOutputs.get(0));
    Random random = new Random();
    PerturbationContext perturbationContext = new PerturbationContext(0L, random, 1);
    LimeConfig limeConfig = new LimeConfig().withSamples(10).withPerturbationContext(perturbationContext);
    LimeExplainer limeExplainer = new LimeExplainer(limeConfig);
    Map<String, Saliency> saliencyMap = limeExplainer.explainAsync(prediction, model).get(Config.INSTANCE.getAsyncTimeout(), Config.INSTANCE.getAsyncTimeUnit());
    for (Saliency saliency : saliencyMap.values()) {
        assertNotNull(saliency);
        List<String> strings = saliency.getTopFeatures(3).stream().map(f -> f.getFeature().getName()).collect(Collectors.toList());
        assertTrue(strings.contains("Actual Speed") || strings.contains("Speed Limit"));
    }
    assertDoesNotThrow(() -> ValidationUtils.validateLocalSaliencyStability(model, prediction, limeExplainer, 1, 0.3, 0.3));
    String decision = "Fine";
    List<PredictionInput> inputs = new ArrayList<>();
    for (int n = 0; n < 10; n++) {
        inputs.add(new PredictionInput(DataUtils.perturbFeatures(predictionInput.getFeatures(), perturbationContext)));
    }
    DataDistribution distribution = new PredictionInputsDataDistribution(inputs);
    int k = 2;
    int chunkSize = 5;
    double f1 = ExplainabilityMetrics.getLocalSaliencyF1(decision, model, limeExplainer, distribution, k, chunkSize);
    AssertionsForClassTypes.assertThat(f1).isBetween(0.5d, 1d);
}

Example 58 with PerturbationContext

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

the class PmmlCompoundScorecardLimeExplainerTest method testExplanationStabilityWithOptimization.

@Test
void testExplanationStabilityWithOptimization() throws ExecutionException, InterruptedException, TimeoutException {
    PredictionProvider model = getModel();
    List<PredictionInput> samples = getSamples();
    List<PredictionOutput> predictionOutputs = model.predictAsync(samples.subList(0, 5)).get();
    List<Prediction> predictions = DataUtils.getPredictions(samples, predictionOutputs);
    long seed = 0;
    LimeConfigOptimizer limeConfigOptimizer = new LimeConfigOptimizer().withDeterministicExecution(true);
    Random random = new Random();
    PerturbationContext perturbationContext = new PerturbationContext(seed, random, 1);
    LimeConfig initialConfig = new LimeConfig().withSamples(10).withPerturbationContext(perturbationContext);
    LimeConfig optimizedConfig = limeConfigOptimizer.optimize(initialConfig, predictions, model);
    assertThat(optimizedConfig).isNotSameAs(initialConfig);
    LimeExplainer limeExplainer = new LimeExplainer(optimizedConfig);
    PredictionInput testPredictionInput = getTestInput();
    List<PredictionOutput> testPredictionOutputs = model.predictAsync(List.of(testPredictionInput)).get(Config.INSTANCE.getAsyncTimeout(), Config.INSTANCE.getAsyncTimeUnit());
    Prediction instance = new SimplePrediction(testPredictionInput, testPredictionOutputs.get(0));
    assertDoesNotThrow(() -> ValidationUtils.validateLocalSaliencyStability(model, instance, limeExplainer, 1, 0.6, 0.6));
}

Example 59 with PerturbationContext

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

the class PmmlCompoundScorecardLimeExplainerTest method testExplanationImpactScoreWithOptimization.

@Test
void testExplanationImpactScoreWithOptimization() throws ExecutionException, InterruptedException, TimeoutException {
    PredictionProvider model = getModel();
    List<PredictionInput> samples = getSamples();
    List<PredictionOutput> predictionOutputs = model.predictAsync(samples.subList(0, 5)).get();
    List<Prediction> predictions = DataUtils.getPredictions(samples, predictionOutputs);
    long seed = 0;
    LimeConfigOptimizer limeConfigOptimizer = new LimeConfigOptimizer().withDeterministicExecution(true).forImpactScore();
    Random random = new Random();
    PerturbationContext perturbationContext = new PerturbationContext(seed, random, 1);
    LimeConfig initialConfig = new LimeConfig().withSamples(10).withPerturbationContext(perturbationContext);
    LimeConfig optimizedConfig = limeConfigOptimizer.optimize(initialConfig, predictions, model);
    assertThat(optimizedConfig).isNotSameAs(initialConfig);
}

Example 60 with PerturbationContext

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

the class PmmlRegressionCategoricalLimeExplainerTest method testExplanationWeightedStabilityWithOptimization.

@Disabled("See KOGITO-6154")
@Test
void testExplanationWeightedStabilityWithOptimization() throws ExecutionException, InterruptedException, TimeoutException {
    PredictionProvider model = getModel();
    List<PredictionInput> samples = getSamples();
    List<PredictionOutput> predictionOutputs = model.predictAsync(samples.subList(0, 5)).get();
    List<Prediction> predictions = DataUtils.getPredictions(samples, predictionOutputs);
    long seed = 0;
    LimeConfigOptimizer limeConfigOptimizer = new LimeConfigOptimizer().withDeterministicExecution(true).withWeightedStability(0.4, 0.6);
    Random random = new Random();
    random.setSeed(seed);
    PerturbationContext perturbationContext = new PerturbationContext(seed, random, 1);
    LimeConfig initialConfig = new LimeConfig().withSamples(10).withPerturbationContext(perturbationContext);
    LimeConfig optimizedConfig = limeConfigOptimizer.optimize(initialConfig, predictions, model);
    assertThat(optimizedConfig).isNotSameAs(initialConfig);
    LimeExplainer limeExplainer = new LimeExplainer(optimizedConfig);
    PredictionInput testPredictionInput = getTestInput();
    List<PredictionOutput> testPredictionOutputs = model.predictAsync(List.of(testPredictionInput)).get(Config.INSTANCE.getAsyncTimeout(), Config.INSTANCE.getAsyncTimeUnit());
    Prediction instance = new SimplePrediction(testPredictionInput, testPredictionOutputs.get(0));
    assertDoesNotThrow(() -> ValidationUtils.validateLocalSaliencyStability(model, instance, limeExplainer, 1, 0.5, 0.7));
}