For further information, please contact the General chair: Pierre-Emmanuel Gaillardon
Professor at Seoul National University, Korea
Towards 4-bit Hardware Accelerator for Very Deep Neural Networks
Low precision is critical in designing hardware accelerators for neural networks. Achieving hardware-friendly sub-8 bit precision for deep networks like ResNet-101 is challenging. In this talk, we introduce a simple quantization method which offers effectively 4-bit precision to deep networks at a negligible accuracy loss (<1% top-1 accuracy loss) and, as a realization of the quantization method, a hardware accelerator called OLAccel (outlier-aware accelerator).
Sungjoo Yoo received Ph.D. from Seoul National University in 2000. He worked as researcher at TIMA laboratory, Grenoble France from 2000 to 2004. He was principal engineer at Samsung System LSI from 2004 to 2008. He was at POSTECH from 2008 to 2015. He joined Seoul National University in 2015 and is now associate professor. His current research interests include low power deep neural networks and near data processing (processing-in-memory and in-storage processing).