Keynote 1: Real-Time Edge AI

Summary form only given, as follows. The complete presentation was not made available for publication as part of the conference proceedings. Advances in neural network revolutionized modern machine intelligence, but important challenges remain when applying these solutions in IoT contexts; specifically, in cost-sensitive applications on lowerend embedded devices. The talk discusses challenges in offering real-time machine intelligence services at the edge to support applications in resource constrained environments. The intersection of IoT applications, real-time requirements, and AI capabilities motivates several important research directions. For example, how to support efficient execution of machine learning components on low-cost edge devices while retaining inference quality and offering confidence estimates in results? How to reduce the need for expensive manual labeling of IoT application data? How to improve the responsiveness of AI components to critical real-time stimuli in their physical environment? How to prioritize and schedule the execution of intelligent data processing workflows on edge-device GPUs? How to exploit data transformations that lead to sparser representations of external physical phenomena to attain more efficient learning and inference? The talk discusses recent advances and presents evaluation results in the context of different real-time edge AI applications.