Image: Dr. In Kee Kim's research team, with international collaborators, has received the Best Paper Award at IEEE EDGE 2024, held from July 7 to 13 in Shenzhen, China. The award-winning paper is titled “Characterizing Deep Learning Model Compression with Post-Training Quantization on Accelerated Edge Devices.” This work is the first comprehensive characterization study of online model compression on resource-constrained edge devices. The team evaluated various deep learning models with different sizes and resource requirements, focusing on post-training quantization (PTQ) using recent NVIDIA edge devices. Their detailed analysis of performance and behavior across different precision modes highlights the challenges and opportunities of PTQ compression methods at the edge.
