activity
20242026
collaborators

6 papers

cs.LG2026

Asymptotic Preservation and Uniform Accuracy of Diffusion and Flow-Matching Samplers

Shiheng Zhang

Diffusion and Gaussian-interpolant flow-matching samplers approach data through a terminal noise floor , a singular limit for manifold-supported or rank-deficient data…

cs.LG2026

Guidance Breaks the Fitted Operator: A Terminal-Fitted Repair for Classifier-Free Guidance

Shiheng Zhang

Classifier-free guidance (CFG) is the standard way to strengthen class-conditioning in diffusion and flow-matching samplers, yet at large guidance it oversaturates and destabilizes…

cs.LG2026

Planner-Admissible Graph-PDE Value Extensions for Sparse Goal-Conditioned Planning

Shiheng Zhang

Sparse goal-conditioned planning with few cost-to-go labels can be viewed as a graph-PDE Dirichlet extension problem: extend sparse labels on a goal-dependent boundary to unlabelle…

stat.ML2025

Low-Rank Evolutionary Deep Neural Networks via Adaptive Tangent-Space Reduction

Jiahao Zhang, Shiheng Zhang, Guang Lin

Evolutionary deep neural networks (EDNNs) solve time-dependent partial differential equations by evolving the neural-network parameters sequentially in time through a local least-s…

math.NA2024

SAV-based entropy-dissipative schemes for a class of kinetic equations

Shiheng Zhang, Jie Shen, Jingwei Hu

We introduce novel entropy-dissipative numerical schemes for a class of kinetic equations, leveraging the recently introduced scalar auxiliary variable (SAV) approach. Both first a…

math.NA2024

Structure preserving schemes for a class of Wasserstein gradient flows

Shiheng Zhang, Jie Shen

We introduce in this paper two time discretization schemes tailored for a range of Wasserstein gradient flows. These schemes are designed to preserve mass, positivity and to be uni…