Towards Green AI-Native Networks: Evaluation of Neural Circuit Policy for Estimating Energy Consumption of Base Stations

NCPs for base station energy consumption estimation aim to optimize radio hardware and AI-based network management, offering reduced computational overhead, energy demand, and robust generalization compared to traditional ML models. They provide a low-cost, scalable solution for green AI-native networks, crucial for sustainable computing in telecommunications. AI and ML algorithms enhance future networks, focusing on energy efficiency and reduced carbon footprint. Deploying efficient models in base stations reduces infrastructure's carbon footprint, addressing challenges in managing diverse, data-distribution-changing models. Sparsely structured neural networks, like NCPs, offer dynamic adaptation during inference, improving generalization and reducing retraining needs. Continuous time networks are suitable for irregularly sampled datasets, valuable for telecommunications. This research proposes small, sparse, energy-efficient neural architectures for the telecom domain, focusing on sustainability goals in AI-native network design.