THURSDAY July 23, 3:30pm - 5:00pm
TOPIC AREA: MACHINE LEARNING/AI, DESIGN
KEYWORD: ARCHITECTURE & SYSTEM DESIGN
EVENT TYPE: RESEARCH REVIEWED
A Stitch in Time Saves AI: Tailoring AI for Hardware Platforms
Chair:
Siva Hari - NVIDIA Corp., Sunnyvale, CA
Co-Chair:
Vikas Chandra - Facebook, Sunnyvale, CA
The papers in this session present approaches to tailor deep neural networks to a wide range of hardware platforms, including RISC-V, GPUs, FPGAs, emerging memory devices and compute-in-storage systems.

88.1Extending the RISC-V ISA for Efficient RNN-Based 5G Radio Resource Management
 Speaker: Renzo Andri - ETH Zurich, Switzerland
 Authors: Renzo Andri - ETH Zurich, Switzerland
Tomas Henriksson - Huawei Technologies Co., Ltd., Kista, Sweden
Luca Benini - ETH Zurich & Univ. of Bologna, Switzerland
88.2Taming Unstructured Sparsity on GPUs via Latency-Aware Optimization
 Speaker: Maohua Zhu - Univ. of California, Santa Barbara, United States
 Authors: Maohua Zhu - Univ. of California, Santa Barbara, United States
Yuan Xie - Univ. of California, Santa Barbara, CA
88.3FTDL: A Tailored FPGA-Overlay for Deep Learning with High Scalability
 Speaker: Runbin Shi - Univ. of Hong Kong, Hong Kong
 Authors: Runbin Shi - Univ. of Hong Kong, Hong Kong
Yuhao Ding - Univ. of Hong Kong, Hong Kong
Xuechao Wei - Peking Univ., Beijing, China
He Li - Univ. of Cambridge, Cambridge, United Kingdom
Hang Liu - Stevens Institute of Technology, NJ
Hayden So - Univ. of Hong Kong, Hong Kong
Caiwen Ding - Univ. of Connecticut, Storrs, CT
88.43D CNN Acceleration on FPGA using Hardware-Aware Pruning
 Speaker: Mengshu Sun - Northeastern Univ. , Boston, MA
 Authors: Mengshu Sun - Northeastern Univ. , Boston, MA
Pu Zhao - Northeastern Univ. , Boston, MA
Mehmet Gungor - Northeastern Univ. , Boston, MA
Miriam Leeser - Northeastern Univ. , Boston, MA
Massoud Pedram - Univ. of Southern California, Los Angeles, CA
Xue Lin - Northeastern Univ. , Boston, MA
88.5STANNIS: Low-Power Acceleration of DNN Training Using Computational Storage Devices
 Speaker: Ali HeydariGorji - Univ. of California, Irvine, CA
 Authors: Ali HeydariGorji - Univ. of California, Irvine, CA
Mahdi Torabzadehkashi - NGD Systems, Inc, CA
Siavash Rezaei - Univ. of California, Irvine, CA
Hossein Bobarshad - NGD Systems, Inc, CA
Vladimir Alves - NGD Systems, Inc, CA
Pai H. Chou - Univ. of California, Irvine, CA
88.6ReSiPE: ReRAM-Based Single-Spiking Processing-In-Memory Engine
 Speaker: Ziru Li - Duke Univ., Durham, NC
 Authors: Ziru Li - Duke Univ., Durham, NC
Bonan Yan - Duke Univ., Durham, NC
Hai Li - Duke Univ., Durham, NC