Adaptive Discovering and Merging for Incremental Novel Class Discovery

TL;DR

This paper introduces Adaptive Discovering and Merging (ADM), a novel framework for incremental class discovery that effectively identifies and integrates new classes while preserving existing knowledge, outperforming previous methods.

Contribution

The paper proposes a new paradigm called ADM with adaptive discovering and merging strategies, including Triple Comparison, Probability Regularization, and a hybrid model structure, to improve incremental class discovery and learning.

Findings

ADM significantly outperforms existing class-incremental NCD methods.

The hybrid model reduces interference and preserves old knowledge.

AMM alleviates catastrophic forgetting in class-IL tasks.

Abstract

One important desideratum of lifelong learning aims to discover novel classes from unlabelled data in a continuous manner. The central challenge is twofold: discovering and learning novel classes while mitigating the issue of catastrophic forgetting of established knowledge. To this end, we introduce a new paradigm called Adaptive Discovering and Merging (ADM) to discover novel categories adaptively in the incremental stage and integrate novel knowledge into the model without affecting the original knowledge. To discover novel classes adaptively, we decouple representation learning and novel class discovery, and use Triple Comparison (TC) and Probability Regularization (PR) to constrain the probability discrepancy and diversity for adaptive category assignment. To merge the learned novel knowledge adaptively, we propose a hybrid structure with base and novel branches named Adaptive Model Merging (AMM), which reduces the interference of the novel branch on the old classes to preserve the previous knowledge, and merges the novel branch to the base model without performance loss and parameter growth. Extensive experiments on several datasets show that ADM significantly outperforms existing class-incremental Novel Class Discovery (class-iNCD) approaches. Moreover, our AMM also benefits the class-incremental Learning (class-IL) task by alleviating the catastrophic forgetting problem.