International Workshop on Highly Efficient Neural Networks Design (HENND)
In conjunction with ESWEEK 2017, Seoul, Korea
October 20, 2017
http://cmalab.snu.ac.kr/HENND2017
Venue: Sapphire Ballroom (3F, Lotte Hotel)
Breakfast and Lunch: Sapphire 1+2+3
Oral presentations: Sapphire 4
Poster: Sapphire Lobby
Call for Paper
The International Workshop on Highly Efficient Neural Networks Design is a forum for presentations of state-of-the-art research in highly efficient neural networks design. The workshop will combine both keynote/invited oral presentations and posters of the papers accepted from regular submissions. Embedded deep learning design contest (EDLDC) is held in conjunction with HENND workshop.
Keynote and oral presentations
Keynote by Prof. Yoshua Bengio (Univ. of Montreal) titled "Towards End-to-End Trainable Hardware"
and invited talks by experts from Baidu, NVIDIA, Samsung Electronics, SK Telecom, DeePhi, ETH Zurich, KAIST, POSTECH and UNIST.
Advance program of oral presentations (PDF) and slides are available now.
Poster paper submission
We invite submissions of poster papers related to (but not limited to) the following topics:
Algorithm optimization (e.g., faster R-CNN, YOLO v2, SSD, etc. for object detection) |
Model compression (e.g., pruning, low rank approximation, etc.) |
Deep learning framework (e.g., Caffe2 or TensorFlow framework for mobile and embedded systems) |
On-device training (e.g., training neural networks on mobile devices like federated learning in Google) |
CPU/DSP/GPU architecture enhancements for neural networks (e.g., matrix multiplication co-processor for neural network acceleration like Tensor Core in V100) |
FPGA/ASIC accelerators for neural networks (e.g., zero-aware low precision CNN/RNN accelerator) |