2026
Quadratic Mixing for the Lee–Sidford Dikin Walk
High dimensional sampling and interior point geometry.
02 · Publications
Selected papers and preprints.
Research across generative inference, scientific machine learning, optimization, probability, and numerical algorithms.
2026
High Dimensional Interpolators Can Be Fragile: Heavy Tails and High Dimensional Large Deviations
Fragility hidden by benign average case behavior.
2026
An AI Assisted Solution to the Signed BAR Conjecture
Uniqueness in the Harrison–Reiman class and a completely S class obstruction.
2026
On the Width Scaling of Neural Optimizers Under Matrix Operator Norms I
Row and column normalization, smoothness, and hyperparameter transfer. Slides
ICML 2026
Simple Unbiased Derivative Free Inference Time Scaling for Diffusion Models via Sequential Monte Carlo on Path Measures
Path space importance sampling and sequential Monte Carlo.
ICML 2026
On the Power of Approximate Reward Models for Inference Time Scaling
Quantifying when approximate rewards can reliably guide additional computation.
ICML 2026
SURGE: Unbiased Data Assimilation for Diffusion Models via Particle Filtering
Particle filtering for unbiased diffusion model data assimilation.
ICLR 2026
Physics Informed Inference Time Scaling via Simulation Calibrated Scientific Machine Learning
Defect correction for high dimensional PDE learning. Slides
2025
Unveiling the Scaling Law of PINNs under Non Euclidean Geometry
A geometric route toward predictable scaling in physics informed neural networks.
2025
What Is a Sketch and Precondition Derivation for Low Rank Approximation?
Inverse power error versus inverse power estimation. Slides
AISTATS 2025
Randomized Iterative Solver as Iterative Refinement: A Simple Fix toward Backward Stability
A stability correction for randomized linear solvers.
ICML 2024
Orthogonal Bootstrap: Efficient Simulation of Input Uncertainty
Sample efficient uncertainty quantification through orthogonalization.
UAI 2024 · Oral
Synthetic Principal Component Design: Fast Covariate Balancing with Synthetic Controls
Spectral structure for balanced experiment design.
ICLR 2024 · Spotlight
Generalization Guarantees of Deep ResNets in the Mean Field Regime
Mean field generalization theory for deep residual networks.
NeurIPS 2023
When Can Regression Adjusted Control Variates Help?
Rare events, Sobolev embedding, and minimax optimality.
ICLR 2023 · Spotlight
Minimax Optimal Kernel Operator Learning via Multilevel Training
Optimal rates for learning infinite dimensional operators. Slides
NeurIPS 2022
Sobolev Acceleration and Statistical Optimality for Learning Elliptic Equations via Gradient Descent
How derivative information changes optimization and statistical rates.
ICLR 2022
Machine Learning for Elliptic PDEs: Fast Rate Generalization Bound, Neural Scaling Law and Minimax Optimality
Sample complexity and optimality for learning PDE solutions. Slides
ICLR 2022
An Unconstrained Layer Peeled Perspective on Neural Collapse
Optimization geometry behind neural collapse.
2022
Importance Tempering: Group Robustness for Overparameterized Models
Learning under group imbalance in overparameterized models.
ICML 2020
A Mean Field Analysis of Deep ResNet and Beyond
Provable optimization through overparameterization from depth.
NeurIPS 2019
You Only Propagate Once: Painless Adversarial Training Using Maximal Principle
Control inspired acceleration for adversarial training.
ICML 2018
PDE Net: Learning PDEs from Data
Learning differential operators and nonlinear dynamics from observations.
ICML 2018
Beyond Finite Layer Neural Networks: Bridging Deep Architectures and Numerical Differential Equations
A differential equation view of deep network architecture.
No publications match this search.