Robust Reconfigurable Intelligent Surfaces via Invariant Risk and Causal\n Representations

2021/05/04 by Sumudu Samarakoon, Jihong Park, Samarakoon, Sumudu +3 · 1 citation
Engineering · #Advanced Wireless Communication Technologies #FOS: Computer and information sciences #FOS: Electrical engineering #Machine Learning (cs.LG) #Modular Robots and Swarm Intelligence #Networking and Internet Architecture (cs.NI) #Robotics and Sensor-Based Localization #Signal Processing (eess.SP) #electronic engineering #information engineering

paper · pdf · doi:10.48550/arxiv.2105.01771

openalex publication_date 2021/05/04 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

In this paper, the problem of robust reconfigurable intelligent surface (RIS)\nsystem design under changes in data distributions is investigated. Using the\nnotion of invariant risk minimization (IRM), an invariant causal representation\nacross multiple environments is used such that the predictor is simultaneously\noptimal for each environment. A neural network-based solution is adopted to\nseek the predictor and its performance is validated via simulations against an\nempirical risk minimization-based design. Results show that leveraging\ninvariance yields more robustness against unseen and out-of-distribution\ntesting environments.\n

Cited by

• Adversarial Training: Enhancing Out-of-Distribution Generalization for Learning Wireless Resource Allocation

Related

• Federated Learning Games for Reconfigurable Intelligent Surfaces via Causal Representations
• Balancing Fairness and Robustness via Partial Invariance
• Unsupervised Invariant Risk Minimization
• Robust Invariant Representation Learning by Distribution Extrapolation
• Learning Optimal Features via Partial Invariance