Lifecycle-Aware Neural Architecture Search

Designing neural networks that are both accurate and efficient has become an urgent challenge as the cost of deploying machine learning systems continues to grow. Traditional Neural Architecture Search (NAS) methods usually optimize for accuracy under constraints such as FLOPs, model size, or latency. However, in practice the lifecycle cost of a model is not only determined by training, but also by inference. While training is a one-off expense, inference is repeated continuously once a model is deployed, and depending on the number of queries served, it can become the dominant contributor to energy use and cost. Recent work has emphasized that inference should be explicitly considered when evaluating efficiency and sustainability of ML models, since deployment often represents the longest and most energy-intensive phase of the lifecycle.

This motivates a lifecycle-aware approach to NAS, where model quality is ensured through accuracy on a held-out test set, while the energy costs of both training and inference are explicitly measured. A lifecycle cost metric can combine training energy with expected inference energy, weighted by the projected number of queries, providing a principled way to compare models across different deployment scenarios. For example, models optimized for low inference cost may offer significant lifetime energy savings in high-usage services, while models with low training cost may be more suitable for rapid prototyping or low-volume applications. Incorporating lifecycle insights into architecture search represents an important step towards aligning ML development with efficiency and environmental goals.

This project would be suitable as a BSc or MSc thesis. If you are interested in machine learning systems and their efficiency in general, this project would be a great fit for you. Depending on the size of the project or thesis (BSc, MSc, etc.) and the number of students in the group, we can adjust the tasks of the project.