Metrics for comparing targets and predictions

[johannes-kk]

Create metrics.cpp and metrics.hpp, similar to losses.* but for comparing targets to predictions.

E.g. accuracy:

double accuracy_score(datavec targets, datavec predictions){
    // todo: verify that target and predict have same size
    double correct = 0;
    for (int i = 0; i < targets.size(); i++){
        if (targets[i] == predictions[i]){
            correct += 1;
        } 
    }
    return correct/targets.size();
}

[wfseaton]

Which metrics would we want? Here is sklearn's list for reference: https://scikit-learn.org/stable/modules/model_evaluation.html

[johannes-kk]

accuracy is all we need to begin with. Precision, Recall, F1-score and the likes would be nice to have, but nothing we need. We're focusing on binary classification for the MVP, and accuracy is plenty to make comparisons between methods and runs.