MSNCIL: A Domain-Agnostic Class-Incremental Learning Method Tailored for Automatic Modulation Recognition

IF 3.7 3区计算机科学 Q2 TELECOMMUNICATIONS

IEEE Communications Letters Pub Date : 2025-04-29 DOI:10.1109/LCOMM.2025.3565561

Zhiwen Deng;Chunbo Luo;Zixi Tang;Yang Luo

引用次数: 0

Abstract

The emergence of new modulation types in 6G challenges the adaptability of deep learning-based automatic modulation recognition (DL-AMR) models. This letter presents multi-state neuron class-incremental learning (MSNCIL), the first domain-agnostic class-incremental learning (CIL) method for AMR. Leveraging the sparsity of wireless signal features, MSNCIL dynamically partitions a DL-AMR model into specialized sub-models, each dedicated to different modulation types. In each session, neurons are selected based on activation values, trained, frozen, and assigned state values. During inference, the session ID of a test sample is identified, which directs the corresponding neurons for recognition. Extensive experiments confirm MSNCIL’s effectiveness.

查看原文本刊更多论文

MSNCIL：一种面向自动调制识别的领域不可知类增量学习方法

6G中新调制类型的出现对基于深度学习的自动调制识别（DL-AMR）模型的适应性提出了挑战。这封信提出了多状态神经元类增量学习（MSNCIL），这是AMR的第一个领域不可知的类增量学习（CIL）方法。利用无线信号特性的稀疏性，MSNCIL动态地将DL-AMR模型划分为专门的子模型，每个子模型专用于不同的调制类型。在每个会话中，根据激活值、训练值、冻结值和分配状态值来选择神经元。在推理过程中，识别测试样本的会话ID，并指导相应的神经元进行识别。大量的实验证实了MSNCIL的有效性。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

IEEE Communications Letters 工程技术-电信学

CiteScore

8.10

自引率

7.30%

发文量

590

审稿时长

2.8 months

期刊介绍： The IEEE Communications Letters publishes short papers in a rapid publication cycle on advances in the state-of-the-art of communication over different media and channels including wire, underground, waveguide, optical fiber, and storage channels. Both theoretical contributions (including new techniques, concepts, and analyses) and practical contributions (including system experiments and prototypes, and new applications) are encouraged. This journal focuses on the physical layer and the link layer of communication systems.