Examples of com.inadco.hbl.math.aggregators.OnlineExpBiasedBinomialSummarizer

com.inadco.hbl.math.aggregators.OnlineExpBiasedBinomialSummarizer

see section 2.2.0 of FAS 2.0 Sale expectaiton.pdf for math details of exponentially weighted biased probability estimator with irregular sampling

This class is introducing conjugate prior bias for binomial summarizer (i.e. x must be 1 or 0, perhaps some observations could be 'partial' but it is not average anymore!)

There's also unbiased version somehwere in mahout (MAHOUT-634 I think) @author dmitriy


    static OnlineExpBiasedBinomialSummarizer createExpBinomialSum(Properties props,
                                                                                   boolean bareInstance) {


        if (bareInstance)
            return new OnlineExpBiasedBinomialSummarizer();


        double dt;
        double m = OnlineExpBiasedBinomialSummarizer.DEFAULT_HISTORY_MARGIN;
        double p0;
        double epsilon = OnlineExpBiasedBinomialSummarizer.DEFAULT_EPSILON;


        if (props.containsKey("m"))
            m = Double.parseDouble(props.getProperty("m"));
        if (props.containsKey("epsilon"))
            epsilon = Double.parseDouble(props.getProperty("epsilon"));


        Validate.isTrue(props.containsKey("dt"), "dt parameter is required for exp binomial summarizer");
        dt = Double.parseDouble(props.getProperty("dt"));


        Validate.isTrue(props.containsKey("p0"), "p0 parameter is required for exp binomial summarizer");
        p0 = Double.parseDouble(props.getProperty("p0"));


        return new OnlineExpBiasedBinomialSummarizer(p0, epsilon, dt, m);
    }

Examples of com.inadco.hbl.math.aggregators.OnlineExpBiasedBinomialSummarizer

Related Classes of com.inadco.hbl.math.aggregators.OnlineExpBiasedBinomialSummarizer