Prototype Based Classification from Hierarchy to Fairness

TL;DR

This paper introduces the concept subspace network (CSN), a versatile neural network architecture that unifies hierarchical and fair classification, enabling flexible model adjustments for different tasks.

Contribution

The paper proposes the CSN architecture, capable of learning multiple concept relationships and transforming between hierarchical and fair classification within a single model.

Findings

CSNs achieve state-of-the-art fair classification results.

CSNs can be transformed into hierarchical classifiers.

CSNs reconcile fairness and hierarchy in one model.

Abstract

Artificial neural nets can represent and classify many types of data but are often tailored to particular applications -- e.g., for "fair" or "hierarchical" classification. Once an architecture has been selected, it is often difficult for humans to adjust models for a new task; for example, a hierarchical classifier cannot be easily transformed into a fair classifier that shields a protected field. Our contribution in this work is a new neural network architecture, the concept subspace network (CSN), which generalizes existing specialized classifiers to produce a unified model capable of learning a spectrum of multi-concept relationships. We demonstrate that CSNs reproduce state-of-the-art results in fair classification when enforcing concept independence, may be transformed into hierarchical classifiers, or even reconcile fairness and hierarchy within a single classifier. The CSN is inspired by existing prototype-based classifiers that promote interpretability.