Heterogeneous Multi-Core Systems for Efficient Edge Machine Learning

Explore a comprehensive lecture on heterogeneous multi-core systems for efficient edge machine learning, delivered by Marian Verhelst at the SPCL_Bcast #43 event on October 26, 2023. Delve into the challenges of embedded ML applications, characterized by diverse workloads including signal processing, GeMM and conv kernels, attention layers, and graph processing. Examine how heterogeneous multicore systems can address accelerator efficiency issues and the associated challenges, such as determining optimal core combinations, efficient workload mapping, and data sharing between cores. Learn about a heterogeneous multi-core system for embedded neural network processing developed at KULeuven MICAS, and gain insights into ongoing research aimed at expanding this system to cover more workloads and heterogeneous cores. Discover valuable information for those interested in edge computing, machine learning optimization, and advanced computer architecture.

[SPCL_Bcast] Heterogeneous multi-core systems for efficient EdgeML