Artificial Neural Networks and Tools for ultra-constrained embedded deployability
Danilo Pau
Technical Director, IEEE & ST Fellow
System Research & Applications
STMicroelectronics
Date: 19 January, 11 am EST – 5pm CET - virtual talk
Duration: 90 min
Zoom Registration: https://mit.zoom.us/meeting/register/tJIvf-2pqzkoEtVVIlO53fGbM-iv2dzpY9Dg
The event is free of charge - open to students.
Is Artificial Intelligence a technology trend or a singularity? Why is tinyML important? Reviewing key AI milestones and learning more about the benefits of tinyML is essential to exceeding the limitations of a centralized Artificial Intelligence approach. To meet the challenges and opportunities in processing data closer to the sensors in real time (AI on the edge), ST proposes a unique 5-step method with a set of related tools to automatically deploy pre-trained neural networks on STM32- and SPC5-based solutions.
Additional considerations on ultra-low-power AI solutions for sensor applications will also be discussed including a more comprehensive AI ecosystem where Hyper Parameter Optimization (HPO) and Neural Network Search (NAS) must have deployable constraints in their inner loops.
This Tech Talk is organized in cooperation with MIT CSAIL.