Tunable Weighted Kernel k-Means for Clustered Cell-Free Networking Acceleration and Beam On-Off Control

Xiankun Zeng, Junyuan Wang, Ke Yue, Miaomiao Dong, Bo Bai

Abstract: Beam-level clustered cell-free networking that partitions beams from multiple base-stations (BSs) and users into non-overlapping subnetworks can enable cooperative transmission among BSs, while avoiding coordinating all the beams at a BS. Previous work adopted spectral clustering to group beams and users into subnetworks, which, however, has cubic complexity and cannot control the number of activated beams. In this paper, a tunable weighted kernel k-means algorithm along with a novel initialization approach and two tuning hyper-parameters are proposed. Simulation results show that the proposed algorithm can successfully accelerate clustered cell-free networking by avoiding eigenvalue decomposition as well as control the number of activated beams by carefully fine-tuning hyper-parameters.

External IDs:dblp:conf/icc/ZengWYD024