3.1. Building a Quantifier

Quantifiers are the main objects of interest in the mlquantify library. Since quantification is still emerging as a research area, new quantifiers are being developed all the time. The library is designed to be extensible, so you can easily add your own quantifiers to test out new ideas, and send them to us for inclusion in the library.

Basic quantifiers are easily implemented by subclassing the Quantifier class. The Quantifier class is an abstract base class that defines the interface for all quantifiers. It provides a number of methods that must be implemented by subclasses, including the fit and predict methods. The fit method is used to train the quantifier on a training set, while the predict method is used to make predictions on a test set.

Warning

Recommended usage

The Quantifier class is not intended to be used directly, and it is recommended that you subclass from AggregativeQuantifier or NonAggregativeQuantifier instead, since these classes provide dynamic handling of binary and multiclass problems and has more functionality.

Note

Method functionality

The fit method should return self to allow for method chaining. And the predictions should be a disctionary containing the predicted proportions for each class. The keys of the dictionary should be the class labels, sorted in ascending order. The values should be the predicted proportions for each class, which should sum to 1.0.

You can implement new quantifiers the following way:

from mlquantify.base import Quantifier

class MyQuantifier(Quantifier):

    def __init__(self, param1, param2):
        self.param1 = param1
        self.param2 = param2

    def fit(self, X, y):
        # Fit the quantifier to the training data
        return self

    def predict(self, X):
        # Make predictions on the test data
        return {0: 0.5, 1: 0.5}  # Example prediction