Neural Architecture Search (NAS) for TinyML | Seminar Lecture скачать в хорошем качестве

Neural Architecture Search (NAS) for TinyML | Seminar Lecture

This session focuses on Neural Architecture Search (NAS) and its role in designing efficient neural networks for TinyML and Edge AI applications. It is part of our TinyML seminar at Mälardalen University (MDU). Unlike other sessions in the series, this lecture is primarily conceptual, as NAS workflows are computationally intensive and time-consuming, making live notebook demonstrations impractical within a seminar setting. Topics covered in this lecture include: • Motivation for NAS in TinyML • Manual design vs automated architecture search • Search spaces, search strategies, and evaluation methods • Hardware-aware NAS for resource-constrained devices • Trade-offs between accuracy, latency, memory, and energy • Challenges of applying NAS in embedded and edge systems This session is intended for students, researchers, and engineers seeking a system-level understanding of architecture optimization for resource-constrained machine learning. 📂 Seminar materials and references: https://github.com/HERO-MDH/TinyML-Se... ▶️ Part of the TinyML Seminar @ MDU playlist.