Advancing Ante-Hoc Explainable Models through Generative Adversarial Networks

TL;DR

This paper introduces a generative adversarial network-based framework that enhances the interpretability and performance of visual classification models by aligning learned concepts with human-interpretable visual properties.

Contribution

It proposes a novel joint training scheme combining explanation generation and adversarial training to produce inherently interpretable deep vision models with semantically meaningful concepts.

Findings

Model produces coherent, human-interpretable concept activations

Learned concepts align with object parts and visual attributes

Adversarial training impacts both classification accuracy and concept quality

Abstract

This paper presents a novel concept learning framework for enhancing model interpretability and performance in visual classification tasks. Our approach appends an unsupervised explanation generator to the primary classifier network and makes use of adversarial training. During training, the explanation module is optimized to extract visual concepts from the classifier's latent representations, while the GAN-based module aims to discriminate images generated from concepts, from true images. This joint training scheme enables the model to implicitly align its internally learned concepts with human-interpretable visual properties. Comprehensive experiments demonstrate the robustness of our approach, while producing coherent concept activations. We analyse the learned concepts, showing their semantic concordance with object parts and visual attributes. We also study how perturbations in the adversarial training protocol impact both classification and concept acquisition. In summary, this work presents a significant step towards building inherently interpretable deep vision models with task-aligned concept representations - a key enabler for developing trustworthy AI for real-world perception tasks.