ACM Transactions on Design Automation of Embedded Systems (TODAES)

Special Issue on Energy-Efficient AI Chips

 

Guest Editors:

  Vikas Chandra, Facebook Reality Labs, vchandra@fb.com

  Yiran Chen, Duke University, yiran.chen@duke.edu

  Sungjoo Yoo, Seoul National University, sungjoo.yoo@gmail.com

  

AI applications opened a new era of computing where 100X better energy-efficient computing is urgently required for edge and server systems. On edge, emerging devices for AR/VR applications, immersive video conferencing, etc., pose challenges in current hardware/software design practices. These applications are realized mostly by AI models, e.g., codec avatar, for both input (e.g., camera images) and output (e.g., graphics rendering) processing. Especially, multi-modality is getting more and more popular for both energy efficiency (e.g., acoustic event triggered vision application) and immersive interaction (e.g., TrueVoice in Google’s Series One). Such emerging edge devices demand extreme energy efficiency and boosted performance to run tens of heavy neural network models in real time and under stringent power budget due to battery and cost reasons. Server is a forerunner in improving energy efficiency by aggressively adopting hardware acceleration. Such a change is essential in supporting ever-increasing model size and training workload in current and future AI model development as well as fast-growing main stream commercial AI models like recommender systems, video analytics, etc.

Topics

We aim at covering state-of-the-art industrial and academic efforts to achieve orders of magnitude better energy efficiency in AI chips, especially, for edge and server systems covering, not limited to, the following topics.

 

  AI chips for energy efficiency

  Hardware accelerator and near-memory processing

  Accelerator architectures and systems for machine learning applications

  Algorithm and hardware co-design for machine learning

  Quantization for low precision

  Model compression

  Neural architecture search

  Neural algorithm optimization for energy efficiency

  Test-time optimization

 

Important Dates

•   Submissions deadline: June 15, 2021

•   First-round review decisions: September 1

•   Deadline for revision submissions: October 1

•   Notification of final decisions: November 15

•   Tentative publication: Spring 2022

 

Submission Information

Authors are encouraged to submit high-quality original research contributions. Please identify clearly the additional material from any original conference or workshop paper in your submitted manuscript. Submissions should be made through the ACM TODAES submission site (http://mc.manuscriptcentral.com/todaes) and formatted according to TODAES author guidelines at: https://dl.acm.org/journal/todaes/author-guidelines.

  

For questions and further information, please contact Sungjoo Yoo/sungjoo.yoo@gmail.com

 

ACM Transactions on Design Automation of Embedded Systems (TODAES) Special Issue on Energy-Efficient AI Chips