Energy and timing constraints change how time-series AI models are chosen
Energy and Latency First: How Deployment Constraints Reverse Time-Series Model Selection
Machine Learning
Summary
Choosing AI models usually focuses on how accurate they are, but other important factors like energy use and speed can be more important when putting these models to real use. The authors studied how different time-series models perform when predicting handovers in cellular networks under strict timing and energy budgets. They found that a model with handcrafted features was more accurate and much more energy-efficient than a popular neural network, which used much more energy and sometimes missed its deadline. This shows that choosing models based only on accuracy can lead to selecting models that don’t work well in real-life conditions.
What this means in practice
- •For cellular network engineers: Deploy time-series models that meet strict energy and timing limits to improve handover predictions in user devices.
- •For embedded system developers: Design energy-efficient time-series inference systems that guarantee prediction before decision deadlines in resource-limited devices.
Authors
Shruti Bothe, Illyyne Saffar, Ayush K. Varshney
Abstract
AI models, and time-series models in particular are typically ranked by accuracy and deployed under constraints absent from that ranking. In this paper, we study a real setting where these constraints are explicit and measurable: handover pre- diction in cellular networks, where inference runs on the User Equipment once per 40 ms measurement period and must complete before the network acts. The timing bound is hard: a late prediction is superseded by the measurement it was predicting, while energy is a cumulative budget spent for the life of the device. Neither appears in the objective these models are selected on. We evaluate eight architectures under a protocol that screens for deadline feasibility at the tail rather than the mean, then ranks the feasible models on measured inference energy, count- ing the feature computation that precedes the forward pass and is usually omitted. Gradient-boosted trees on handcrafted temporal features are the most accurate model (F1=0.974) and among the cheapest to run, while the neural architecture an accuracy-only benchmark would rank first (CNN-1D, F1=0.914) costs over 3 times more energy per inference under a realistic duty-cycled deployment pattern. We give three conditions under which this outcome should be expected, so the finding applies beyond telecommunications. Accuracy-only benchmarking does not merely under-report deployment cost. It can rank first a model that cannot be deployed.