p11 / Conference
Carbon-Aware Machine Learning: A Case Study on Cellular Traffic Forecasting with Spiking Neural Networks
Artificial Intelligence Applications and Innovations (AIAI 2024), 178-191 · 2024
In plain language
Can spiking neural networks reduce the energy cost of forecasting cellular traffic without sacrificing prediction quality?
Research context
It connects bio-inspired modeling with energy-conscious decisions about operational machine learning.
Approach
Spiking models are compared with convolutional networks and multilayer perceptrons using measurements from three Barcelona locations.
Findings
The study reports lower energy costs with spiking models while maintaining forecast quality in its evaluated setting.
Summary based on abstract.