F2: ML at the Extreme Edge: Machine Learning as the Killer IoT App
2020
Provides an abstract of the tutorial presentation and may include a brief professional biography of the presenter. The complete presentation was not made available for publication as part of the conference proceedings. This forum puts a clear focus on machine learning in very low power edge devices (μW to mW), rather than focusing on its use in data centers. While GPUs for training in data centers, and large-scale inference engines are the common case today, the combination of IoT and ML brings new capabilities to edge devices. Speakers will motivate this area and provide insights on the impact of resource constraints on both training and inference in such devices. In addition, key application areas are discussed in depth, namely audio and imaging. The role of new memory-centric design approaches in edge-based ML is also discussed. Finally, the software environment is discussed, to make hardware designers aware of the opportunities for power-constrained ML algorithm implementations.