Kaiwen Wu

PhD Student

Computer and Information Science
University of Pennsylvania

Email: kaiwenwu@seas.upenn.edu

How to pronounce my first name

Curriculum Vitae  /  GitHub  /  Google Scholar  /  Twitter  /  Blog

About Me

I am a final-year PhD student in the Department of Computer and Information Science at the University of Pennsylvania, where I am advised by Jacob Gardner. Before coming to Penn, I completed my MMath degree in Computer Science at the University of Waterloo.

I am interested in machine learning and optimization. My recent work focuses on scaling up computation in probabilistic machine learning. Specifically, I work on Gaussian processes, variational inference, and Bayesian optimization. I am also interested in convex optimization and deep generative modeling.

I am looking for full-time positions in industry beginning in 2026.

Research

* indicates equal contribution.

Publications

• Mixed Likelihood Variational Gaussian Processes
  Kaiwen Wu, Craig Sanders, Benjamin Letham and Phillip Guan
  arXiv preprint arXiv:2503.04138 (2025)
  [paper] [code]


• Computation-Aware Gaussian Processes: Model Selection And Linear-Time Inference
  Jonathan Wenger, Kaiwen Wu, Philipp Hennig, Jacob R. Gardner, Geoff Pleiss and John P. Cunningham
  Advances in Neural Information Processing Systems (NeurIPS 2024)
  [paper] [code]


• Understanding Stochastic Natural Gradient Variational Inference
  Kaiwen Wu and Jacob R. Gardner
  International Conference on Machine Learning (ICML 2024)
  [paper] [code]


• Large-Scale Gaussian Processes via Alternating Projection
  Kaiwen Wu, Jonathan Wenger, Haydn Jones, Geoff Pleiss and Jacob R. Gardner
  International Conference on Artificial Intelligence and Statistics (AISTATS 2024)
  [paper] [code]


• The Behavior and Convergence of Local Bayesian Optimization
  Kaiwen Wu, Kyurae Kim, Roman Garnett and Jacob R. Gardner
  Advances in Neural Information Processing Systems (NeurIPS 2023)
  Spotlight Presentation
  [paper] [code]


• On the Convergence of Black-Box Variational Inference
  Kyurae Kim, Jisu Oh, Kaiwen Wu, Yian Ma and Jacob R. Gardner
  Advances in Neural Information Processing Systems (NeurIPS 2023)
  [paper] [code]


• Variational Gaussian Processes with Decoupled Conditionals
  Xinran Zhu, Kaiwen Wu, Natalie Maus, Jacob Gardner and David Bindel
  Advances in Neural Information Processing Systems (NeurIPS 2023)
  [paper] [code]


• Practical and Matching Gradient Variance Bounds for Black-Box Variational Bayesian Inference
  Kyurae Kim, Kaiwen Wu, Jisu Oh and Jacob R Gardner
  International Conference on Machine Learning (ICML 2023)
  Oral Presentation
  [paper] [code]


• Local Bayesian Optimization via Maximizing Probability of Descent
  Quan Nguyen*, Kaiwen Wu*, Jacob R. Gardner and Roman Garnett
  Advances in Neural Information Processing Systems (NeurIPS 2022)
  Oral Presentation
  [paper] [code]


• Discovering Many Diverse Solutions with Bayesian Optimization
  Natalie Maus, Kaiwen Wu, David Eriksson and Jacob Gardner
  International Conference on Artificial Intelligence and Statistics (AISTATS 2022)
  Notable Paper Award
  [paper] [code]


• Stronger and Faster Wasserstein Adversarial Attacks
  Kaiwen Wu, Allen Houze Wang and Yaoliang Yu
  International Conference on Machine Learning (ICML 2020)
  [paper] [code]


• On Minimax Optimality of GANs for Robust Mean Estimation
  Kaiwen Wu, Gavin Weiguang Ding, Ruitong Huang and Yaoliang Yu
  International Conference on Artificial Intelligence and Statistics (AISTATS 2020)
  [paper] [code]

Workshop Papers

• A Fast, Robust Elliptical Slice Sampling Implementation for Linearly Truncated Multivariate Normal Distributions
  Kaiwen Wu and Jacob R. Gardner
  Workshop on Bayesian Decision-Making and Uncertainty at NeurIPS 2024
  [paper] [code]


• Newton-type Methods for Minimax Optimization
  Guojun Zhang, Kaiwen Wu, Pascal Poupart and Yaoliang Yu
  Workshop on Beyond First-Order Methods in ML Systems at ICML 2021
  [paper] [code]

Miscellaneous

The following notes are for self-reference only. Some notes take forever to finish, unfortunately.

• A Gentle Introduction to Variational Gaussian Processes
• Concepts in Smooth Manifolds and Differential Geometry
• Normalization in Gaussian Processes
• L1 Norm Projection

I have reviewed (or will review) for the following conferences: AAAI 2021, AISTATS 2021, ICML 2023, NeurIPS 2023, ICLR 2024, AISTATS 2024, ICML 2024, NeurIPS 2024, ICLR 2025, ICML 2025, NeurIPS 2025, AAAI 2026, ICLR 2026.

I have reviewed (or will review) for the following journals: TMLR, JMLR.

A website calculating an upper bound of the Erdős number. My Erdős number is 3.