Xiuyu Li

News

• [01/2022] TorchSparse accepted to MLSys 2022
• [12/2021] Recieved an honorable mention on CRA 2022 Outstanding Undergraduate Researcher Award
• [12/2021] I graduated from Cornell!
• [09/2021] One paper accepted to NeurIPS 2021
• [03/2021] One paper accepted to WWW 2021 GLB
• [08/2020] Attended the Cornell, Maryland, and Max Planck Pre-Doctoral Research School (CMMRS) 2020

Publications

TorchSparse: Efficient Point Cloud Inference Engine
Haotian Tang*, Zhijian Liu*, Xiuyu Li*, Yujun Lin, Song Han
Conference on Machine Learning and Systems (MLSys), 2022
[abs]  [draft]  [code]

Coming soon.

Large Scale Learning on Non-Homophilous Graphs: New Benchmarks and Strong Simple Methods
Derek Lim*, Felix Hohne*, Xiuyu Li*, Sijia Linda Huang, Vaishnavi Gupta, Omkar Bhalerao, Ser-Nam Lim
Advances in Neural Information Processing Systems (NeurIPS), 2021
Previous version: New Benchmarks for Learning on Non-Homophilous Graphs
The Web Conference (WWW) Workshop on Graph Learning Benchmarks, 2021
[arXiv]  [pdf]  [code & datasets]
[abs]  [arXiv]  [pdf]  [code & datasets]

Many widely used datasets for graph machine learning tasks have generally beenhomophilous, where nodes with similar labels connect to each other. Recently, new Graph Neural Networks (GNNs) have been developed that move beyond the homophily regime; however, their evaluation has often been conducted on small graphs with limited application domains. We collect and introduce diverse non-homophilous datasets from a variety of application areas that have up to 384x more nodes and 1398x more edges than prior datasets. We further show that existing scalable graph learning and graph minibatching techniques lead to performance degradation on these non-homophilous datasets, thus highlighting the need for further work on scalable non-homophilous methods. To address these concerns, we introduce LINKX — a strong simple method that admits straightforward minibatch training and inference. Extensive experimental results with representative simple methods and GNNs across our proposed datasets show that LINKX achieves state-of-the-art performance for learning on non-homophilous graphs.

Experience

Massachusetts Institute of Technology

Research Intern
June 2021 - Present

Worked with Prof. Song Han on efficient machine learning algorithms and systems co-design for 3D point cloud data.

National University of Singapore

Research Intern
May 2019 - July 2019

Worked with Prof. Wei Wang on building SINGA-Auto, a distributed system that trains and deploys machine learning models using AutoML.

Projects

• Quantizing Lottery Tickets
  CS 6787 Advanced Machine Learning Systems
  We proposed a novel model compression scheme that applies post-training quantization on subnetworks derived by iterative magnitude prunning, which reduces model parameters by 83%-96% and number of bits from 32 to 6-8 without loss of accuracy on MNIST/CIFAR10 datasets.
  [report] [code]
• Community Detection in Networks by Identifying Causality
  CS 6241 Numerical Methods for Data Science
  We proposed a novel approach to to find community structure, and relationships between communities in network-generated data by identifying causality with sparse plus low-rank matrix factorization and conditional causation entropy.
  [report]
• Open World Active Learning
  We proposed a system that can actively discover and learn new classes in the wild through the interaction with an oracle for image recognition, and open-sourced a library that contains basic to state-of-the-art active learning and open set recognition algorithms.
  [code]
• Continual Learning with Lottery Tickets
  We proposed and demonstrated the effectiveness of a novel training scheme that derives subnetworks in a model for each task through iterative magnitude prunning to resist catastrophic forgetting, a phenomena in which a model overfits to the most recently seen data in a multi-task learning setting.
  [report]

Activities

Cornell University Artificial Intelligence

Co-President
June 2021 - Present

Undergraduate Researcher
Sept 2019 - Present

Teaching

Teaching Assistant:

• Fall 2021, Fall 2020: CS 4780 Machine Learning for Intelligent Systems
• Summer 2021: SoNIC Summer Research Workshop
  • This workshop has a goal to increase exposure and enhance research capabilities for underrepresented minorities.
  • Research Lab Notebooks
• Spring 2021: CS 4787 Principles of Large-Scale Machine Learning
• Spring 2020: CS 2802 Discrete Structures - Honors

Misc

• My {surname + given name} in Chinese: 李岫宇
• Chat with me!
• I love playing archery (recurve bow) in my free time.