Graduate Thesis Or Dissertation


Graph Imitation Learning for Optimal Joint Beamforming and Antenna Selection Public Deposited

Downloadable Content

Download PDF


Attribute NameValues
  • Transmit beamforming is an important technique employed to improve efficiency and signal quality in wireless communication systems by steering signals towards their in- tended users. It often arises jointly with the antenna selection problem due to various reasons, such as limited number of radio frequency (RF) chains and energy/resource effi- ciency considerations. The joint robust beamforming and antenna selection (RBF&AS) problem is a mixed integer nonlinear program. Due to the NP-hard combinatorial nature of this problem, majority of existing methods rely on various heuristics, e.g., continuous approximations, greedy search, and supervised machine learning. However, these heuris- tics do not guarantee the optimality (or even feasibility) of the considered problem. To address this issue, we design an effective branch-and-bound (B&B) based method that guarantees optimal solutions to the problem of interest. To avoid the potentially costly nature of the proposed B&B algorithm, a machine learning-based scheme is pro- posed that expedites the B&B search by skipping intermediate steps of the algorithm. The learning model is based on a graph neural network (GNN) that provides resilience to commonly encountered problems in wireless comunications, namely, the change of problem size (e.g., the number of users) across the training and test stages. Finally, we provide a comprehensive theoretical analysis, which shows the proposed GNN-based method can reduce the complexity of the B&B method while retaining global optimality under reasonable conditions. Extensive numerical simulations show that the proposed method can provide near-optimal solution with an order-of-magnitude speedup relative to the B&B.
Resource Type
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Committee Member
Academic Affiliation
Rights Statement
Funding Statement (additional comments about funding)
  • This work is supported by National Science Foundation (NSF) under Project CNS-2003082. Their work is also supported by a gift from Intel through the MLWiNS program.
Peer Reviewed



This work has no parents.

In Collection: