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.

On The Efficient Design Of Neural Networks In Communication Systems

Optimizing Quantized Neural Networks in a Weak Curvature Manifold

Modular Neural Network-Based Models of High-Speed Link Transceivers

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

Quantization of Neural Network Equalizers in Optical Fiber Transmission Experiments

Learning to Optimize: Training Deep Neural Networks for Interference Management

Learn to Communicate with Neural Calibration: Scalability and Generalization

Joint Transceiver Optimization for Wireless Communication PHY with Convolutional Neural Network

Learning Low Resource Consumption CNN through Pruning and Quantization

Online Deep Neural Network for Optimization in Wireless Communications

Low-precision deep-learning-based automatic modulation recognition system

Joint Transceiver Optimization for Wireless Communication PHY Using Neural Network.

Residual Quantization for Low Bit-Width Neural Networks.

Joint Neural Network Equalizer and Decoder

An Interpretable Mapping from a Communication System to a Neural Network for Optimal Transceiver-Joint Equalization

Quantization Networks

Low-bitwidth Convolutional Neural Networks for Wireless Interference Identification

Efficient Residual Shrinkage CNN Denoiser Design for Intelligent Signal Processing: Modulation Recognition, Detection, and Decoding

Frequency-Domain Optimization of Signal Waveform Shaper in High-Speed Optical Communication Systems Considering ADC Quantization Noise

A 4-Bit Integer-Only Neural Network Quantization Method Based on Shift Batch Normalization

Bit Efficient Quantization for Deep Neural Networks