Optimizers

A BaseOptimizer object is all you need in order to perform a singleRegression on a set of events.

Based on the specific distribution you want to use to fit your data you can instantiate one of three BaseOptimizer objects:

• auto opt = InverseGaussianOptimizer()

• auto opt = GaussianOptimizer()

• auto opt = LogNormalOptimizer()

Once you have instantiated your optimizer, you have to build a WeightsProducer instance, a PointProcessDataset and call the .singleRegression() method as you can see in the example below:

#include "pointprocess/WeightsProducer.h"
#include "pointprocess/PointProcessDataset.h"
#include "pointprocess/optimizers/LogNormalOptimizer.h"


int main()
{
    // Retrieve some data...
    auto td = getTestData();
    // Since td.testEvents is a std::vector<double>, in this case we need to copy it inside a std::deque<double>
    // in order to build a PointProcessDataset.
    std::deque<double> events;
    for (int i = 75; i < 300; i++){
        events.push_back(td.testEvents[i]);
    }

    auto wp = WeightsProducer(1.0);

    auto dataset = PointProcessDataset::load(
            events, // event_times
            9, // AR_ORDER
            true, // hasTheta0
            wp
            );

    auto optimizer = LogNormalOptimizer();
    auto result = optimizer.singleRegression(
            dataset,
            false, // rightCensoring
            10000 // maxIter
            );
}

The output of the singleRegression() method is a RegressionResult:

struct RegressionResult{
    double theta0;
    Eigen::VectorXd thetaP;
    double mu;
    double sigma;
    double lambda;
    double meanInterval;
    double time = 0.0;
    unsigned long nIter;
    double likelihood;
    bool eventHappened = false;
}