Automatic Concept Embedding Model (ACEM): No train-time concepts, No issue!

TL;DR

ACEM introduces an automatic approach to learn concept embeddings in neural networks, eliminating the need for manual concept annotations and enhancing interpretability without sacrificing performance.

Contribution

This paper presents ACEM, a novel concept embedding model that automatically learns concept annotations, reducing annotation costs and improving explainability in neural networks.

Findings

ACEM effectively learns concepts without manual annotations.

ACEM maintains high performance while improving interpretability.

The approach reduces the need for costly data labeling.

Abstract

Interpretability and explainability of neural networks is continuously increasing in importance, especially within safety-critical domains and to provide the social right to explanation. Concept based explanations align well with how humans reason, proving to be a good way to explain models. Concept Embedding Models (CEMs) are one such concept based explanation architectures. These have shown to overcome the trade-off between explainability and performance. However, they have a key limitation -- they require concept annotations for all their training data. For large datasets, this can be expensive and infeasible. Motivated by this, we propose Automatic Concept Embedding Models (ACEMs), which learn the concept annotations automatically.