This Probably Looks Exactly Like That: An Invertible Prototypical Network

TL;DR

This paper introduces ProtoFlow, an invertible, concept-based neural network combining generative flow models with prototypes, enhancing interpretability and performance in supervised learning.

Contribution

It proposes ProtoFlow, a novel invertible model that integrates flow-based generative models with prototypes, improving interpretability and predictive accuracy.

Findings

Sets new state-of-the-art in generative and predictive modeling

Achieves comparable predictive performance to existing prototypical networks

Enables richer, more robust interpretation of prototypes

Abstract

We combine concept-based neural networks with generative, flow-based classifiers into a novel, intrinsically explainable, exactly invertible approach to supervised learning. Prototypical neural networks, a type of concept-based neural network, represent an exciting way forward in realizing human-comprehensible machine learning without concept annotations, but a human-machine semantic gap continues to haunt current approaches. We find that reliance on indirect interpretation functions for prototypical explanations imposes a severe limit on prototypes' informative power. From this, we posit that invertibly learning prototypes as distributions over the latent space provides more robust, expressive, and interpretable modeling. We propose one such model, called ProtoFlow, by composing a normalizing flow with Gaussian mixture models. ProtoFlow (1) sets a new state-of-the-art in joint generative and predictive modeling and (2) achieves predictive performance comparable to existing prototypical neural networks while enabling richer interpretation.