Jingfeng Wu | 吴京风

I am a postdoc fellow at the Simons Institute at UC Berkeley hosted by Peter Bartlett and Bin Yu. I am a part of the NSF/Simons Collaboration on the Theoretical Foundations of Deep Learning.

I obtained my Ph.D. in Computer Science at Johns Hopkins University, advised by Vladimir Braverman, and B.S. and M.S. at Peking University. I was selected as a 2023 Rising Star in Data Science by UChicago and UCSD.

Email | CV | arXiv | Google Scholar

Research

My research focuses on bridging the gap between theory and practice in machine learning by developing efficient algorithms to solve real-world problems and providing a deep understanding of the underlying theoretical principles.

I work on the theory and algorithms of deep learning, and related topics in algorithms, machine learning, optimization, and statistical learning theory.

News

[02/2025] Organizing a deep learning theory workshop at Simons.
[01/2025] One paper accepted to ICLR 2025.

Prior to 2025

[09/2024] Three papers accepted to NeurIPS 2024; huge congrats to Yuhang, Licong, and Ruiqi!!

[05/2024] One paper accepted to COLT 2024.

[03/2024] Invited talk at UCLA CS.

[02/2024] Invited talk at UC Berkeley Biostatistics.

[01/2024] Two papers accepted to ICLR 2024.

[10/2023] Selected as a 2023 Rising Star in Data Science by UChicago and UCSD.

[09/2023] Two papers accepted to NeurIPS 2023.

[06/2023] Defended my PhD dissertation!

[05/2023] One paper accepted to CoLLAs 2023; congrats to Haoran!

[04/2023] One paper accepted to ICML 2023.

[03/2023] I will join the Simons Institute at UC Berkeley as a postdoc.

[09/2022] Two papers accepted to NeurIPS 2022.

[06/2022] Interning at Google Research Seattle during the summer of 2022.

[05/2022] One paper accepted to ICML 2022 as long presentation!

[01/2022] One paper accepted to AISTATS 2022.

[12/2021] Passed the PhD candidacy exam.

[09/2021] Two papers accepted to NeurIPS 2021.

[05/2021] Awarded MINDS Summer Data Science Fellowship!

[05/2021] One paper accepted to COLT 2021.

[02/2021] In a relationship with Yuan, happy Valentine's day~

[01/2021] One paper is accepted to ICLR 2021.

[10/2020] One paper will be presented at Workshop on Optimization for Machine Learning.

[05/2020] Two papers accepted to ICML 2020.

[11/2019] One paper will be presented at the Workshop on Optimization for Machine Learning.

[09/2019] Starting working at Hopkins. Veritas vos liberabit!

[06/2019] Graduated from Peking University.

[04/2019] One paper accepted to ICML 2019.

[03/2019] One paper accepted to CVPR 2019 as an oral presentation.

[12/2018] Looking for a Ph.D. position in machine learning starting in fall 2019.

Publications

* indicates equal contribution, † indicates alphabetical order.

Selected Papers

Benefits of Early Stopping in Gradient Descent for Overparameterized Logistic Regression
JW, Peter L. Bartlett^†, Matus Telgarsky^†, Bin Yu^†
arXiv 2025 | slides
Scaling Laws in Linear Regression: Compute, Parameters, and Data
Licong Lin, JW, Sham M. Kakade, Peter L. Bartlett, Jason D. Lee
NeurIPS 2024
Large Stepsize Gradient Descent for Non-Homogeneous Two-Layer Networks: Margin Improvement and Fast Optimization
Yuhang Cai, JW, Song Mei, Michael Lindsey, Peter L. Bartlett
NeurIPS 2024
Large Stepsize Gradient Descent for Logistic Loss: Non-Monotonicity of the Loss Improves Optimization Efficiency
JW, Peter L. Bartlett^†, Matus Telgarsky^†, Bin Yu^†
COLT 2024 | slides | poster
How Many Pretraining Tasks Are Needed for In-Context Learning of Linear Regression?
JW, Difan Zou, Zixiang Chen, Vladimir Braverman, Quanquan Gu, Peter L. Bartlett
ICLR 2024 (spotlight) | slides | poster
Implicit Bias of Gradient Descent for Logistic Regression at the Edge of Stability
JW, Vladimir Braverman, Jason D. Lee
NeurIPS 2023 (spotlight) | slides | poster
Finite-Sample Analysis of Learning High-Dimensional Single ReLU Neuron
JW*, Difan Zou*, Zixiang Chen*, Vladimir Braverman, Quanquan Gu, Sham M. Kakade
ICML 2023 | slides | poster
The Power and Limitation of Pretraining-Finetuning for Linear Regression under Covariate Shift
JW*, Difan Zou*, Vladimir Braverman, Quanquan Gu, Sham M. Kakade
NeurIPS 2022 | slides | poster
Last Iterate Risk Bounds of SGD with Decaying Stepsize for Overparameterized Linear Regression
JW*, Difan Zou*, Vladimir Braverman, Quanquan Gu, Sham M. Kakade
ICML 2022 (long presentation) | slides | poster
The Benefits of Implicit Regularization from SGD in Least Squares Problems
Difan Zou*, JW*, Vladimir Braverman, Quanquan Gu, Dean P. Foster, Sham M. Kakade
NeurIPS 2021
Benign Overfitting of Constant-Stepsize SGD for Linear Regression
Difan Zou*, JW*, Vladimir Braverman, Quanquan Gu, Sham M. Kakade
COLT 2021 (journal version in JMLR 2023) | slides

All Preprints

Benefits of Early Stopping in Gradient Descent for Overparameterized Logistic Regression
JW, Peter L. Bartlett^†, Matus Telgarsky^†, Bin Yu^†
arXiv 2025 | slides
Context-Scaling versus Task-Scaling in In-Context Learning
Amirhesam Abedsoltan, Adityanarayanan Radhakrishnan, JW, Mikhail Belkin
arXiv 2024

All Refereed Papers

How Does Critical Batch Size Scale in Pre-training?
Hanlin Zhang, Depen Morwani, Nikhil Vyas, JW, Difan Zou, Udaya Ghai, Dean Foster, Sham Kakade
ICLR 2025
Large Stepsize Gradient Descent for Non-Homogeneous Two-Layer Networks: Margin Improvement and Fast Optimization
Yuhang Cai, JW, Song Mei, Michael Lindsey, Peter L. Bartlett
NeurIPS 2024
Scaling Laws in Linear Regression: Compute, Parameters, and Data
Licong Lin, JW, Sham M. Kakade, Peter L. Bartlett, Jason D. Lee
NeurIPS 2024
In-Context Learning of a Linear Transformer Block: Benefits of the MLP Component and One-Step GD Initialization
Ruiqi Zhang, JW, Peter L. Bartlett
NeurIPS 2024
Large Stepsize Gradient Descent for Logistic Loss: Non-Monotonicity of the Loss Improves Optimization Efficiency
JW, Peter L. Bartlett^†, Matus Telgarsky^†, Bin Yu^†
COLT 2024 | slides | poster
How Many Pretraining Tasks Are Needed for In-Context Learning of Linear Regression?
JW, Difan Zou, Zixiang Chen, Vladimir Braverman, Quanquan Gu, Peter L. Bartlett
ICLR 2024 (spotlight) | slides | poster
Risk Bounds of Accelerated SGD for Overparameterized Linear Regression
Xuheng Li, Yihe Deng, JW, Dongruo Zhou, Quanquan Gu
ICLR 2024
Implicit Bias of Gradient Descent for Logistic Regression at the Edge of Stability
JW, Vladimir Braverman, Jason D. Lee
NeurIPS 2023 (spotlight) | slides | poster
Private Federated Frequency Estimation: Adapting to the Hardness of the Instance
JW, Wennan Zhu, Peter Kairouz, Vladimir Braverman
NeurIPS 2023 | slides | poster
Fixed Design Analysis of Regularization-Based Continual Learning
Haoran Li*, JW*, Vladimir Braverman
CoLLAs 2023
Finite-Sample Analysis of Learning High-Dimensional Single ReLU Neuron
JW*, Difan Zou*, Zixiang Chen*, Vladimir Braverman, Quanquan Gu, Sham M. Kakade
ICML 2023 | slides | poster
The Power and Limitation of Pretraining-Finetuning for Linear Regression under Covariate Shift
JW*, Difan Zou*, Vladimir Braverman, Quanquan Gu, Sham M. Kakade
NeurIPS 2022 | slides | poster
Risk Bounds of Multi-Pass SGD for Least Squares in the Interpolation Regime
Difan Zou*, JW*, Vladimir Braverman, Quanquan Gu, Sham M. Kakade
NeurIPS 2022
Last Iterate Risk Bounds of SGD with Decaying Stepsize for Overparameterized Linear Regression
JW*, Difan Zou*, Vladimir Braverman, Quanquan Gu, Sham M. Kakade
ICML 2022 (long presentation) | slides | poster
Gap-Dependent Unsupervised Exploration for Reinforcement Learning
JW, Vladimir Braverman, Lin F. Yang
AISTATS 2022 | slides | poster
The Benefits of Implicit Regularization from SGD in Least Squares Problems
Difan Zou*, JW*, Vladimir Braverman, Quanquan Gu, Dean P. Foster, Sham M. Kakade
NeurIPS 2021
Accommodating Picky Customers: Regret Bound and Exploration Complexity for Multi-Objective Reinforcement Learning
JW, Vladimir Braverman, Lin F. Yang
NeurIPS 2021 | slides | poster
Lifelong Learning with Sketched Structural Regularization
Haoran Li, Aditya Krishnan, JW, Soheil Kolouri, Praveen K. Pilly, Vladimir Braverman
ACML 2021
Benign Overfitting of Constant-Stepsize SGD for Linear Regression
Difan Zou*, JW*, Vladimir Braverman, Quanquan Gu, Sham M. Kakade
COLT 2021 (journal version in JMLR 2023) | slides
Direction Matters: On the Implicit Bias of Stochastic Gradient Descent with Moderate Learning Rate
JW, Difan Zou, Vladimir Braverman, Quanquan Gu
ICLR 2021 | slides | poster
Obtaining Adjustable Regularization for Free via Iterate Averaging
JW, Vladimir Braverman, Lin F. Yang
ICML 2020 | slides | code
On the Noisy Gradient Descent that Generalizes as SGD
JW, Wenqing Hu, Haoyi Xiong, Jun Huan, Vladimir Braverman, Zhanxing Zhu
ICML 2020 | slides | code
Tangent-Normal Adversarial Regularization for Semi-supervised Learning
Bing Yu*, JW*, Jinwen Ma, Zhanxing Zhu
CVPR 2019 (oral) | slides | poster | code
The Anisotropic Noise in Stochastic Gradient Descent: Its Behavior of Escaping from Minima and Regularization Effects
Zhanxing Zhu*, JW*, Bing Yu, Lei Wu, Jinwen Ma
ICML 2019 | slides | poster | code

Invited Talks

GD for Logistic Regression: Benefits of Early Stopping
- [02/2025] UC Berkeley, Deep Learning Theory Workshop, hosted by Peter Bartlett et al.
Reimagining Gradient Descent: Large Stepsize, Oscillation, and Acceleration
- [01/2025] UCLA, Level Set Meeting, hosted by Shu Liu, Stanley Osher
- [09/2024] Simons Foundation, MoDL Annual Meeting, hosted by Peter Bartlett, Rene Vidal
- [05/2024] UC San Diego, MoDL Collaboration Meeting, hosted by Chaoyue Liu et al.
- [03/2024] UCLA, Computer Science Seminar, hosted by Quanquan Gu
- [02/2024] UC Berkeley, Biostatistics Seminar, hosted by Lexin Li
- [02/2024] UC San Diego, Group Seminar, hosted by Mikhail Belkin
New Insights about SGD: Stepsize, Risk Convergence, and Implicit Regularization
- [10/2023] UC Davis, Statistics Seminar, hosted by Xiao Hui Tai
Implicit Bias of Gradient Descent for Logistic Regression at the Edge of Stability
- [05/2023] TTIC, MoDL Collaboration Meeting, hosted by Sam Buchanan et al.
The Implicit Regularization of SGD in Least Square Problems and Beyond
- [01/2023] Rice University, Algorithms and ML Seminar, hosted by Anastasios Kyrillidis
- [12/2022] Princeton University, Group Seminar, hosted by Jason D. Lee
- [11/2022] Georgia Tech, Group Seminar, hosted by Molei Tao
- [08/2022] Google Research, Learning Theory Seminar, hosted by Mehryar Mohri
- [06/2022] MPI & UCLA, Math Machine Learning Seminar, hosted by Guido Montufar

Services

Conference Reviewer
ICML (2020 - 2025), NeurIPS (2020 - 2024), ICLR (2021 - 2025), AISTATS (2021 - 2023), UAI (2023)
Conference Program Committee Member
AAAI (2021 - 2023)
Journal Reviewer
JMLR, TPAMI, TMLR, SIMODS, JAIR, Applied Probability Journals, IEEE Transactions on Information Theory,

People

Family

Recent Student Collaborators

Yuhang Cai Licong Lin Ruiqi Zhang

Collaborators

Amir Abedsoltan Peter L. Bartlett Mikhail Belkin Vladimir Braverman Zixiang Chen Yihe Deng Udaya Ghai Quanquan Gu Peter Kairouz Sham M. Kakade Jason D. Lee Haoran Li Xuheng Li Michael Lindsey Song Mei Depen Morwani Adit Radhakrishnan Matus Telgarsky Nikhil Vyas Lin F. Yang Bin Yu Hanlin Zhang Wennan Zhu Dongruo Zhou Difan Zou