About

I am a 5th year PhD candidate at Cornell University in the Department of Computer Science, where I am advised by Prof. Chris De Sa. I also work closely with Prof. Vitaly Shmatikov at Cornell Tech. I received my bachelor degree in Mathematics from Shanghai Jiao Tong University in July 2019, where I am fourtunate to work with Prof. John E. Hopcroft and Huan Long.

For my research, I am intrigued by the prospect of integrating data geometry into machine learning and NLP, as it helps capture diverse properties exhibited by data across various tasks. I'm also dedicated to developing efficient algorithmic and library solutions to ensure the robust numerical computation of low-precision ML models. Additionally, my interests extend to LLMs, machine learning privacy and robustness, along with a curiosity for emerging cognitive learning paradigms such as vector symbolic architectures and hyperdimensional computing.

I am actively looking for a postdoc position. Please feel free to contact me if you have any opportunities or if you would like to discuss potential collaborations.

[Curriculum Vitae] [Google Scholar] [Github] [LinkedIn]

Publications

Manuscripts

Tao Yu, Congzheng Song, Jianyu Wang, Mona Chitnis. "Momentum Approximation in Asynchronous Private Federated Learning", (Under Review)

Tao Yu, Yichi Zhang, Zhiru Zhang, Christopher De Sa. "FedHDC: Secure and Private Federated Hyperdimensional Computing", (To Appear)

Tao Yu, Eugene Bagdasaryan, Vitaly Shmatikov. "Salvaging Federated Learning by Local Adaptation", [Code].


Conferences

Tao Yu, Gaurav Gupta, Karthick Gopalswamy, Amith R Mamidala, Hao Zhou, Jeffrey Huynh, Youngsuk Park, Ron Diamant, Anoop Deoras, Luke Huan. "Collage: Light-Weight Low-Precision Strategy for LLM Training". In 41th International Conference on Machine Learning (ICML 2024).

Tao Yu*, Toni J.B. Liu*, Albert Tseng, Christopher De Sa. "Shadow Cones: Unveiling Partial Orders in Hyperbolic Space", [Code]. In 12th International Conference on Learning Representations (ICLR 2024)

Albert Tseng, Tao Yu, Toni J.B. Liu, Christopher De Sa. "Coneheads: Hierarchy Aware Attention", [Code]. In 37th Conference on Neural Information Processing Systems (NeurIPS 2023)

Tao Yu, Christopher De Sa. "Random Laplacian Features For Learning with Hyperbolic Space" [Arxiv version], [ICLR version], [Code]. In 11th International Conference on Learning Representations (ICLR 2023)

Tao Yu*, Yichi Zhang*, Zhiru Zhang, Christopher De Sa. "Understanding Hyperdimensional Computing for Parallel Single-Pass Learning", [Code]. In 36th Conference on Neural Information Processing Systems (NeurIPS 2022)

Tao Yu*, Wentao Guo*, Jianan Canal Li*, Tiancheng Yuan*, Christopher De Sa. "MCTensor: A High-Precision Deep Learning Library with Multi-Component Floating-Point", [Code]. In 39th International Conference on Machine Learning (ICML 2022), Workshop on Hardware Aware Efficient Training (HAET 2022)

Tao Yu, Christopher De Sa. "Representing Hyperbolic Space Accurately using Multi-Component Floats". In 35th Conference on Neural Information Processing Systems (NeurIPS 2021)

Tao Yu, Christopher De Sa. "Numerically Accurate Hyperbolic Embeddings Using Tiling-Based Models", Spotlight, [Compression Code, Learning Code, Poster]. In 33rd Conference on Neural Information Processing Systems (NeurIPS 2019)

Tao Yu*, Shengyuan Hu*, Chuan Guo, Wei-Lun Chao, Kilian Q. Weinberger. "A New Defense Against Adversarial Images: Turning a Weakness into a Strength", [Code, Poster]. In 33rd Conference on Neural Information Processing Systems (NeurIPS 2019)

Felix Wu*, Tianyi Zhang*, Amauri Holanda de Souza Jr.*, Christopher Fifty, Tao Yu, Kilian Q. Weinberger. "Simplifying Graph Convolutional Networks", [Code]. In 36th International Conference on Machine Learning (ICML 2019).

Tao Yu, Huan long, John Hopcroft. "Curvature-based Comparison of Two Neural Networks". In 24th International Conference on Pattern Recognition (ICPR 2018).

Education & Experience

Research Intern June 2020 - Aug. 2022

It's my fortune to intern at Apple MLPT Privacy team working with experienced reserachers and engineers including Ulfar Erlingsson, Vojta Jina, Martin Pelikan, Omid Javidbakht and etc., to look into some topics on federated learning and ml privacy.

Research Intern July 2018 - Dec. 2018

Happy to get the research intern opportunity in Cornell CS from Prof. Kilian Q. Weinberger, to work on defenses against adversarial examples and simplifying GCN for NLP tasks. I also work closely with Prof. Chris De Sa on developing numerically robust and accurate models for hyperbolic embeddings of graphs.

B.S. in Mathematics Sep. 2015 - July 2019

It's my great honor to major in Mathematics and Applied Mathematics (ZhiYuan honours programme) at Shanghai Jiao Tong University, where I am so lucky to work with Prof. John E. Hopcroft and Huan Long, we analyzed both theoretically and experimentally of the intrinsic dimension of the manifolds embedded in neural networks.

Recent Projects



Professional Activities

PC/Reviewer

ICML, NeurIPS, ICLR, AISTATS, MLSys, KDD, SDM.

Talks

NeurIPS 2019, "Numerically Accurate Hyperbolic Embeddings Using Tiling-Based Models", Slides.
VSAONLINE 2022, "Understanding hyperdimensional computing for parallel single-pass learn- ing", Video, Slides.
Workshop on Privacy Preserving Machine Learning 2024, "Efficient Asynchronous Private Federated Learning".