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

Example 71 with Value

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

the class PartialDependencePlotExplainer method explainFromDataDistribution.

private List<PartialDependenceGraph> explainFromDataDistribution(PredictionProvider model, int outputSize, DataDistribution dataDistribution) throws InterruptedException, ExecutionException, TimeoutException {
    long start = System.currentTimeMillis();
    List<PartialDependenceGraph> pdps = new ArrayList<>();
    List<FeatureDistribution> featureDistributions = dataDistribution.asFeatureDistributions();
    // fetch entire data distributions for all features
    List<PredictionInput> trainingData = dataDistribution.sample(config.getSeriesLength());
    // create a PDP for each feature
    for (FeatureDistribution featureDistribution : featureDistributions) {
        // generate (further) samples for the feature under analysis
        // TBD: maybe just reuse trainingData
        List<Value> xsValues = featureDistribution.sample(config.getSeriesLength()).stream().sorted(// sort alphanumerically (if Value#asNumber is NaN)
        Comparator.comparing(Value::asString)).sorted(// sort by natural order
        (v1, v2) -> Comparator.comparingDouble(Value::asNumber).compare(v1, v2)).distinct().collect(Collectors.toList());
        List<Feature> featureXSvalues = // transform sampled Values into Features
        xsValues.stream().map(v -> FeatureFactory.copyOf(featureDistribution.getFeature(), v)).collect(Collectors.toList());
        // create a PDP for each feature and each output
        for (int outputIndex = 0; outputIndex < outputSize; outputIndex++) {
            PartialDependenceGraph partialDependenceGraph = getPartialDependenceGraph(model, trainingData, xsValues, featureXSvalues, outputIndex);
            pdps.add(partialDependenceGraph);
        }
    }
    long end = System.currentTimeMillis();
    LOGGER.debug("explanation time: {}ms", (end - start));
    return pdps;
}

Example 72 with Value

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

the class PartialDependencePlotExplainer method collapseMarginalImpacts.

/**
 * Collapse value counts into marginal impacts.
 * For numbers ({@code Type.NUMBER.equals(type))} this is just the average of each value at each feature value.
 * For all other types the final {@link Value} is just the most frequent.
 *
 * @param valueCounts the frequency of each value at each position
 * @param type the type of the output
 * @return the marginal impacts
 */
private List<Value> collapseMarginalImpacts(List<Map<Value, Long>> valueCounts, Type type) {
    List<Value> yValues = new ArrayList<>();
    if (Type.NUMBER.equals(type)) {
        List<Double> doubles = valueCounts.stream().map(v -> v.entrySet().stream().map(e -> e.getKey().asNumber() * e.getValue() / config.getSeriesLength()).mapToDouble(d -> d).sum()).collect(Collectors.toList());
        yValues = doubles.stream().map(Value::new).collect(Collectors.toList());
    } else {
        for (Map<Value, Long> item : valueCounts) {
            long max = 0;
            String output = null;
            for (Map.Entry<Value, Long> entry : item.entrySet()) {
                if (entry.getValue() > max) {
                    max = entry.getValue();
                    output = entry.getKey().asString();
                }
            }
            yValues.add(new Value(output));
        }
    }
    return yValues;
}

Example 73 with Value

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

the class PmmlRegressionCategoricalLimeExplainerTest method getModel.

private PredictionProvider getModel() {
    return inputs -> CompletableFuture.supplyAsync(() -> {
        List<PredictionOutput> outputs = new ArrayList<>();
        for (PredictionInput input1 : inputs) {
            List<Feature> features1 = input1.getFeatures();
            CategoricalVariablesRegressionExecutor pmmlModel = new CategoricalVariablesRegressionExecutor(features1.get(0).getValue().asString(), features1.get(1).getValue().asString());
            PMML4Result result = pmmlModel.execute(categoricalVariableRegressionRuntime);
            String score = result.getResultVariables().get("result").toString();
            PredictionOutput predictionOutput = new PredictionOutput(List.of(new Output("result", Type.NUMBER, new Value(score), 1d)));
            outputs.add(predictionOutput);
        }
        return outputs;
    });
}

Example 74 with Value

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

the class RemotePredictionProvider method toPredictionOutput.

protected PredictionOutput toPredictionOutput(JsonObject mainObj) {
    if (mainObj == null || !mainObj.containsKey("result")) {
        LOG.error("Malformed json {}", mainObj);
        return null;
    }
    List<Output> resultOutputs = toOutputList(mainObj.getJsonObject("result"));
    List<String> resultOutputNames = resultOutputs.stream().map(Output::getName).collect(toList());
    Map<String, TypedValue> mappedOutputs = predictionOutputs.stream().collect(Collectors.toMap(HasNameValue::getName, HasNameValue::getValue));
    // It's possible that some outputs are missing in the response from the prediction service
    // (e.g. when the generated perturbed inputs don't make sense and a decision is skipped).
    // The explainer, however, may throw exceptions if it can't find all the inputs that were
    // specified in the execution request.
    // Here we take the outputs received from the prediction service and we fill (only if needed)
    // the missing ones with Output objects containing "null" values of type UNDEFINED, to make
    // the explainer happy.
    List<Output> outputs = Stream.concat(resultOutputs.stream().filter(output -> mappedOutputs.containsKey(output.getName())), mappedOutputs.keySet().stream().filter(key -> !resultOutputNames.contains(key)).map(key -> new Output(key, Type.UNDEFINED, new Value(null), 1d))).collect(toList());
    return new PredictionOutput(outputs);
}

Example 75 with Value

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

the class PmmlRegressionLimeExplainerTest method getModel.

private PredictionProvider getModel() {
    return inputs -> CompletableFuture.supplyAsync(() -> {
        List<PredictionOutput> outputs = new ArrayList<>();
        for (PredictionInput input1 : inputs) {
            List<Feature> features1 = input1.getFeatures();
            LogisticRegressionIrisDataExecutor pmmlModel = new LogisticRegressionIrisDataExecutor(features1.get(0).getValue().asNumber(), features1.get(1).getValue().asNumber(), features1.get(2).getValue().asNumber(), features1.get(3).getValue().asNumber());
            PMML4Result result = pmmlModel.execute(logisticRegressionIrisRuntime);
            String species = result.getResultVariables().get("Species").toString();
            double score = Double.parseDouble(result.getResultVariables().get("Probability_" + species).toString());
            PredictionOutput predictionOutput = new PredictionOutput(List.of(new Output("species", Type.TEXT, new Value(species), score)));
            outputs.add(predictionOutput);
        }
        return outputs;
    });
}