Sai Zhang

About Me

 

I am a research scientist at Meta Reality Lab. Previously I was a postdoctoral fellow at Harvard University, and I received my PhD degree of Computer Science at the same university. Before coming to Harvard, I received my bachelor's degree and master's degree in Electrical Engineering and Statistics from University of Toronto. My research interest lies in algorithm/hardware codesign for efficient deep neural network (DNN) implementation. I am also interested in Multi-Agent Reinforcement Learning (MARL) and its application.

 

My chinese name is 张赛骞

 

zhangs@g.harvard.edu

Selected Publications

A Multi-agent Reinforcement Learning Approach for Efficient Client Selection in Federated Learning

 

Sai Qian Zhang, Jieyu Lin, Qi Zhang

Thirty-sixth AAAI Conference on Artificial Intelligence (AAAI), 2022.

(Paper)

FAST: DNN Training Under Variable Precision Block Floating Point with Stochastic Rounding

 

Sai Qian Zhang, Bradley McDanel, H.T. Kung

28th IEEE International Symposium on High-Performance Computer Architecture (HPCA), 2022

(Paper)

Training for Multi-resolution Inference Using Reusable Quantization Terms

 

Sai Qian Zhang, Bradley McDanel, H.T. Kung, Xin Dong

26th ACM International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS), 2021

(Paper)(Code)

Succinct and Robust Multi-Agent Communication With Temporal Message Control  

 

Sai Qian Zhang, Jieyu Lin, Qi Zhang

Neural Information Processing System (NeurIPS), 2020

(Paper)(Code)

Term Quantization: Furthering Quantization at Run Time

 

Sai Qian Zhang*, H.T. Kung*, Bradley McDanel* (equal contribution)

ACM/IEEE International Conference for High Performance Computing, Networking,

Storage and Analysis (SC), 2020

(Paper)

Adaptive Distributed Convolutional Neural Network Inference at the Network Edge with ADCNN

 

Sai Qian Zhang, Qi Zhang, Jieyu Lin

International Conference on Parallel Processing (ICPP), 2020

(Paper)(Video)

Efficient Communication in Multi-Agent Reinforcement Learning via Variance Based Control

 

Sai Qian Zhang, Qi Zhang, Jieyu Lin

Neural Information Processing System (NeurIPS), 2019

(Paper)(Code)

Packing Sparse Convolutional Neural Networks for Efficient Systolic Array Implementations: Column Combining Under Joint Optimization

 

H.T. Kung*, Bradley McDanel*, Sai Qian Zhang* (equal contribution with alphabetical author list)

26th ACM International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS), 2019

(Paper)(Code)

Full-stack Optimization for Accelerating CNNs with FPGA Validation

 

Sai Qian Zhang*, Bradley McDanel*, H.T. Kung, Xin Dong (equal contribution)

ACM International Conference on Supercomputing (ICS), 2019

(Paper)(Code)

News

[3/2023] My high school mentee, Gavin an, has successfully finished his AI project on efficient reinforcement learning. A paper is submitted to JEI.

 

[9/2022] Start working at Meta!

 

[7/2022] Our paper “Hyperspherical Federated Learning" is accepted by ECCV 2022!

 

[7/2022] I gave an invited talk at AI times on multi-agent reinformcent learning and its applications.

 

[6/2022] I gave an invited talk at IEEE Dallas Circuits and Systems Conference (DCAS), 2022.

 

[4/2022] I started my AI memtorship at Veritas AI!

 

[2/2022] I started my postdoc study at Harvard!

 

[12/2021] I successfully defended my PhD!  

 

[11/2021] Our paper “Learning Advanced Client Selection Strategy for Federated Learning" is accepted by AAAI 2022!

 

[10/2021] Our paper “FAST: DNN Training Under Variable Precision Block Floating Point with Stochastic Rounding" is accepted by IEEE HPCA 2022!

 

[06/2021] Finished my internship at Microsoft, such a great place for research!

 

[03/2021] I started my virtual internship at Microsoft Research, Redmond.

 

[03/2021] I gave a guest lecture on DNN accelerator design on Havard Course ES201, hosted by Prof. Demba Ba.

 

[01/2021] One paper got accepted by IEEE International Symposium on Circuits & Systems (ISCAS), 2021.

 

[12/2020] I presented (virtually) our work "Succinct and Robust Multi-Agent Communication With Temporal Message Control" in NeurIPS'20.

 

[11/2020] I presented (virtually) our work "Term quantization: furthering quantization at run time" in SC'20.

 

[11/2020] Our paper “Training for Multi-resolution Inference Using Reusable Quantization Terms" is accepted by ACM ASPLOS 2021!

 

[09/2020] Our paper "Succinct and Robust Multi-Agent Communication With Temporal Message Control" is accepted by NeurIPS 2020!

 

[08/2020] I presented our work "Adaptive Distributed Convolutional Neural Network Inference at the Network Edge with ADCNN" in ICPP 2020.

 

[06/2020] Our paper "Term Revealing: Furthering Quantization at Run Time on Quantized DNNs" is accepted by ACM/IEEE SC 2020!

 

[05/2020] One paper "Adaptive Distributed Convolutional Neural Network Inference at the Network Edge with ADCNN" is accepted by ACM ICPP 2020!

 

[02/2020] One paper is accepted by IEEE Symposium on Security and Privacy (S&P) Deep Learning and Security workshop, 2020.

 

[12/2019] I attended NeurIPS and presented our work in Vancouver, Canada. 

 

[11/2019] Our paper "RTN: Reparameterized Ternary Network" is accepted by AAAI, 2020!