Temporal Imbalance of Positive and Negative Supervision in Class-Incremental Learning

TL;DR

This paper identifies temporal imbalance as a key factor causing prediction bias in class-incremental learning and proposes a novel loss function, TAL, to address it, leading to improved long-term stability.

Contribution

The paper introduces the concept of temporal imbalance in CIL and proposes the Temporal-Adjusted Loss (TAL) to dynamically reweight supervision, enhancing model stability.

Findings

TAL reduces forgetting in CIL benchmarks.

TAL effectively mitigates prediction bias caused by temporal imbalance.

Theoretical analysis confirms TAL's effectiveness under imbalance conditions.

Abstract

With the widespread adoption of deep learning in visual tasks, Class-Incremental Learning (CIL) has become an important paradigm for handling dynamically evolving data distributions. However, CIL faces the core challenge of catastrophic forgetting, often manifested as a prediction bias toward new classes. Existing methods mainly attribute this bias to intra-task class imbalance and focus on corrections at the classifier head. In this paper, we highlight an overlooked factor -- temporal imbalance -- as a key cause of this bias. Earlier classes receive stronger negative supervision toward the end of training, leading to asymmetric precision and recall. We establish a temporal supervision model, formally define temporal imbalance, and propose Temporal-Adjusted Loss (TAL), which uses a temporal decay kernel to construct a supervision strength vector and dynamically reweight the negative supervision in cross-entropy loss. Theoretical analysis shows that TAL degenerates to standard cross-entropy under balanced conditions and effectively mitigates prediction bias under imbalance. Extensive experiments demonstrate that TAL significantly reduces forgetting and improves performance on multiple CIL benchmarks, underscoring the importance of temporal modeling for stable long-term learning.