Global vs Local#

Objectives: what you will take away#

  • How-To retrieve global and local metrics.

Prerequisites: before you begin#

Data#

Our example dataset for this recipe is the well known Adult dataset. It is accessible via the pmlb package installed earlier. We use the fetch_data() function to retrieve the dataset in Step 1 below.

Concepts & Terminology#

How-To Guide#

Setup#

The user guide assumes you have create and setup a Trainee as demonstrated in basic workflow. The Trainee will be referenced as trainee in the sections below.

Global metrics#

Global metrics in Howso refers to calculations done using all of the cases available. Sometimes they are sampled for efficiency, however they are still representative of the overall Trainee.

Global metrics in Howso are calculated internally using a leave one out approach to the datapoints trained into the trainee that is called by the react_aggregate() method.

# Getting global feature contributions
feature_contributions_robust = trainee.react_react_aggregate(
    action_feature=action_features[0],
    details={"feature_contributions_robust": True}
)

Local metrics#

Local metrics in Howso refers to calculations done using all of the cases available. Local metrics are calculated using the local space of the provided case(s). These cases may either be new cases or existing cases. The local space refers to the set of closest cases to the provided case(s). While the exact number of cases that consists of the local space varies depending on several factors, generally it includes at least 30 cases if there is enough cases.

Local metrics are controlled through the details parameter in Trainee.react().

details = {'feature_contributions_robust' : True}
# Getting global feature contributions
results = trainee.react(
    test_case[context_features],
    context_features=context_features,
    action_features=action_features,
    details=details
)

feature_contributions = results['details']['feature_contributions_robust']

API References#