Abstract:In recent years, the application of neural networks (NNs) in wireless communication has garnered widespread attention and proven successful. However, conventional learning-based NNs often suffer from poor calibration, meaning that they struggle to reliably quantify prediction confidence and lack proper uncertainty estimation. This limitation becomes especially critical for next generation communication systems, particularly in complex industrial scenarios with stringent reliability requirements. Previous efforts to enhance model calibration have primarily centered on modifying NNs’ training processes. However, these methods often demand significant computing resources, making them impractical for resource-constrained scenarios. In this paper, we investigate a distributed wireless communication system involving multiple users and propose a novel approach to improve model calibration. Our method focuses on enhancing calibration during the inference stage of NNs by introducing a power control mechanism. Notably, existing research indicates that many NNs exhibit overconfidence, i.e., the NN’s confidence exceeds its actual accuracy. Leveraging this insight, we exploit the inherent noise and fading in wireless systems to naturally reduce the NN’s confidence while preserving accuracy. To achieve this, we employ linear relaxation-based perturbation analysis (LiRPA) to approximate the relationship between the perturbed output and the input perturbation of the NN. Subsequently, we devise an optimization problem by leveraging the analyzed relationship and the definition of perfect calibration. It is aimed at finding the input perturbation that maximizes the probability of the model achieving perfect calibration. Finally, considering different channel conditions and a given specific modulation method, we derive the optimal transmission power based on bit error rate (BER) formula. Simulation results demonstrate that our proposed power control method exhibits significant advantages in model calibration compared to several traditional approaches.

A Power-Only Fast Calibration Method for Phased Array Using Convolution Neural Network

A Fast Calibration Method For Phased Arrays By Using The Graph Coloring Theory

Calibration of a Digital Phased Array by Using NCO Phase Increasing Algorithm

Gain and Phase Autocalibration of Large Uniform Rectangular Arrays for Underwater 3-D Sonar Imaging Systems

An Online Calibration Method Using Hadamard–Fourier Clustering and Neural Network for Large-Scale Phased Arrays

Multi-Element Phased Array Calibration Method by Solving Linear Equations

Gain and Phase Calibration of Uniform Rectangular Arrays Based on Convex Optimization and Neural Networks

An Efficient Near-Field Model Gain-Phase Self-Calibration Method for Uniform Rectangular Arrays

Gain and Phase Calibration of Sensor Arrays for Underwater 3-D Sonar Imaging System Based on TSI Direction Finding

Phased Array Antenna with Self-Calibration Network and Improved Scanning Performance

Study on a Fast Measurement Method of Phased Array Antennas

A High-Resolution and Low-Frequency Acoustic Beamforming Based on Bayesian Inference and Non-Synchronous Measurements

An Improved Complex Signal-Based Calibration Method for Beam-Steering Phased Array

Equivalent Magnetic Dipole Matrix Reconstruction Method Based on Deep Learning for Magnitude-Only Near-Field Scanning

A Method for Active Three-Dimensional Array Calibration

Self-supervised Calibration Method of Array Antenna for High-Precision GNSS Application

A Self-Adaptive Phased Array Conformed on a Highly Flexible Aperture

Study on a Fast Measurement and Identifying Malfunction Elements in Phased Array Antennas

ROM-Based Real-Time Analysis of Electromagnetic Performance for APAA in a Digital Twin System

A Sensing-Based Over-the-Air Phased Array Antennas Phase Calibration Method With Greedy Algorithm

Power Control for NN-based Wireless Distributed Inference with Improved Model Calibration