Balancing the Causal Effects in Class-Incremental Learning

TL;DR

This paper identifies causal imbalance as a key issue in class-incremental learning and proposes BaCE, a method to balance causal effects from new and old data, improving continual learning performance across tasks.

Contribution

The paper introduces BaCE, a novel causal balancing approach that mitigates interference between new and old class data in incremental learning.

Findings

BaCE outperforms existing CIL methods on image, text, and NER tasks.

Causal imbalance is a critical factor in catastrophic forgetting.

Balancing causal effects improves model adaptation to all classes.

Abstract

Class-Incremental Learning (CIL) is a practical and challenging problem for achieving general artificial intelligence. Recently, Pre-Trained Models (PTMs) have led to breakthroughs in both visual and natural language processing tasks. Despite recent studies showing PTMs' potential ability to learn sequentially, a plethora of work indicates the necessity of alleviating the catastrophic forgetting of PTMs. Through a pilot study and a causal analysis of CIL, we reveal that the crux lies in the imbalanced causal effects between new and old data. Specifically, the new data encourage models to adapt to new classes while hindering the adaptation of old classes. Similarly, the old data encourages models to adapt to old classes while hindering the adaptation of new classes. In other words, the adaptation process between new and old classes conflicts from the causal perspective. To alleviate this problem, we propose Balancing the Causal Effects (BaCE) in CIL. Concretely, BaCE proposes two objectives for building causal paths from both new and old data to the prediction of new and classes, respectively. In this way, the model is encouraged to adapt to all classes with causal effects from both new and old data and thus alleviates the causal imbalance problem. We conduct extensive experiments on continual image classification, continual text classification, and continual named entity recognition. Empirical results show that BaCE outperforms a series of CIL methods on different tasks and settings.