A Fine-Grained Approach to Explaining Catastrophic Forgetting of Interactions in Class-Incremental Learning

Xu Cheng, Hao Zhang, Zechao Li

Keywords: Interpretability, catastrophic forgetting, interaction

TL;DR: This paper explains catastrophic forgetting and unifies different class-incremental learning methods from a novel perspective of interactions.

Abstract: This paper explains catastrophic forgetting in class incremental learning (CIL) from a novel perspective of interactions (non-linear relationship) between different input variables. Specifically, we make the first attempt to explicitly identify and quantify which interactions w.r.t. previous classes that are forgotten and preserved over incremental steps, and reveal their distinct behaviors, so as to provide a more fine-grained explanation of catastrophic forgetting. Based on the forgotten interactions, we provide a unified explanation for the effectiveness of different CIL methods in mitigating catastrophic forgetting, i.e., these methods all reduce the forgetting of interactions w.r.t. previous classes, particularly those of low complexities, although these methods are originally designed based on different intuitions and observations. Intrigued by this, we further propose a simple-yet-efficient method with theoretical guarantees to investigate the role of low-complexity interactions in the resistance of catastrophic forgetting, and discover that low-order interaction serves as an effective factor in resisting catastrophic forgetting. The code will be released if the paper is accepted.

Primary Area: interpretability and explainable AI

Submission Number: 6425