InfoCons: Identifying Interpretable Critical Concepts in Point Clouds via Information Theory

TL;DR

This paper introduces InfoCons, an information-theoretic framework for identifying interpretable, causally relevant critical concepts in point clouds, enhancing model interpretability in safety-critical applications like autonomous vehicles.

Contribution

We propose a novel method, InfoCons, that decomposes point clouds into meaningful concepts using information theory, enabling causal and human-understandable explanations of model predictions.

Findings

Outperforms baseline methods on synthetic datasets

Scalable and flexible to real-world datasets

Effective in applications requiring critical concept scores

Abstract

Interpretability of point cloud (PC) models becomes imperative given their deployment in safety-critical scenarios such as autonomous vehicles. We focus on attributing PC model outputs to interpretable critical concepts, defined as meaningful subsets of the input point cloud. To enable human-understandable diagnostics of model failures, an ideal critical subset should be *faithful* (preserving points that causally influence predictions) and *conceptually coherent* (forming semantically meaningful structures that align with human perception). We propose InfoCons, an explanation framework that applies information-theoretic principles to decompose the point cloud into 3D concepts, enabling the examination of their causal effect on model predictions with learnable priors. We evaluate InfoCons on synthetic datasets for classification, comparing it qualitatively and quantitatively with four baselines. We further demonstrate its scalability and flexibility on two real-world datasets and in two applications that utilize critical scores of PC.