Quantum Fusion Learning for 6G Networks With Kolmogorov–Arnold Networks

Abstract

Driven by the vision of intelligent and adaptive connectivity in the sixth-generation (6G) networks, future wireless systems are expected to support advanced applications such as cognitive resource allocation, semantic communication (SC), and low-latency integrated sensing and communication (ISAC) or digital twins. However, realizing these capabilities requires learning frameworks capable of efficiently modeling complex, high-dimensional, and dynamic communication environments. To address this challenge, we propose a Kolmogorov-Arnold network (KAN)-enhanced quantum fusion learning (QFL) model designed to advance intelligent wireless communication systems. Specifically, we introduce a hybrid quantum-classical (HQC) architecture that integrates parametrized quantum circuits with deep learning (DL) models based on residual networks (ResNets) and shifted window transformers (SwinTs), further augmented with KAN layers. This HQC framework potentially addresses the limitations of conventional DL (ResNets/SwinTs) by accurately modeling nonlinear relationships, capturing high-order dependencies, and flexibly adapting to dynamic wireless environments. We evaluate the proposed QFL framework on three representative 6G tasks: 1) satellite remote-sensing scene classification (RSSCN7 dataset), 2) ISAC beamforming prediction (DeepSense 6G dataset)—which supports cognition-level resource allocation and low-latency ISAC—and 3) generative SC classification (GenSC-6G dataset)—relevant to semantic environment modeling and digital twin updates. Numerical results demonstrate that the designed QFL architecture delivers competitive performance, underscoring its potential for enabling intelligent and secure 6G systems.