Weijie Huang

Weijie Huang, Zhiping Li

A mixed finite element method combining an iso-parametric $Q_2$-$P_1$ element and an iso-parametric $P_2^+$-$P_1$ element is developed for the computation of multiple cavities in incompressible nonlinear elasticity. The method is analytically proved to be locking-free and convergent, and it is also shown to be numerically accurate and efficient by numerical experiments. Furthermore, the newly developed accurate method enables us to find an interesting new bifurcation phenomenon in multi-cavity growth.

Leyi Zhao, Weijie Huang, Yitong Guo et al.

Optimizing scientific computing algorithms for modern GPUs is a labor-intensive and iterative process involving repeated code modification, benchmarking, and tuning across complex hardware and software stacks. Recent work has explored large language model (LLM)-assisted evolutionary methods for automated code optimization, but these approaches primarily rely on outcome-based selection and random mutation, underutilizing the rich trajectory information generated during iterative optimization. We propose PhyloEvolve, an LLM-agent system that reframes GPU-oriented algorithm optimization as an In-Context Reinforcement Learning (ICRL) problem. This formulation enables trajectory-conditioned reuse of optimization experience without model retraining. PhyloEvolve integrates Algorithm Distillation and prompt-based Decision Transformers into an iterative workflow, treating sequences of algorithm modifications and performance feedback as first-class learning signals. To organize optimization history, we introduce a phylogenetic tree representation that captures inheritance, divergence, and recombination among algorithm variants, enabling backtracking, cross-lineage transfer, and reproducibility. The system combines elite trajectory pooling, multi-island parallel exploration, and containerized execution to balance exploration and exploitation across heterogeneous hardware. We evaluate PhyloEvolve on scientific computing workloads including PDE solvers, manifold learning, and spectral graph algorithms, demonstrating consistent improvements in runtime, memory efficiency, and correctness over baseline and evolutionary methods. Code is published at: https://github.com/annihi1ation/phylo_evolve

Weijie Huang, Xinran Ruan

We propose a sharp-interface model for solid-state dewetting of thin films with wetting potential, where the wetting effect is incorporated through a thickness-dependent surface energy. The model is governed by surface diffusion together with natural boundary conditions, and describes the morphological evolution of the film-vapor interface. For its numerical approximation, we develop an efficient semi-implicit finite element method based on a Taylor expansion of the wetting-potential term. Numerical simulations in two dimensions show that the proposed model and method can capture various dewetting phenomena. They also indicate that, as the range of the wetting potential tends to zero, the proposed model approaches the sharp-interface model with thickness-independent surface energy proposed in [1]. The model and numerical method are further extended to three dimensions, where the computations capture complex morphological evolution in solid-state dewetting.

Pufan Zou, Shijia Zhao, Weijie Huang et al.

Recently, Visual Foundation Models (VFMs) have shown a remarkable generalization performance in 3D perception tasks. However, their effectiveness in large-scale outdoor datasets remains constrained by the scarcity of accurate supervision signals, the extensive noise caused by variable outdoor conditions, and the abundance of unknown objects. In this work, we propose a novel label-free learning method, Adaptive Label Correction (AdaCo), for 3D semantic segmentation. AdaCo first introduces the Cross-modal Label Generation Module (CLGM), providing cross-modal supervision with the formidable interpretive capabilities of the VFMs. Subsequently, AdaCo incorporates the Adaptive Noise Corrector (ANC), updating and adjusting the noisy samples within this supervision iteratively during training. Moreover, we develop an Adaptive Robust Loss (ARL) function to modulate each sample's sensitivity to noisy supervision, preventing potential underfitting issues associated with robust loss. Our proposed AdaCo can effectively mitigate the performance limitations of label-free learning networks in 3D semantic segmentation tasks. Extensive experiments on two outdoor benchmark datasets highlight the superior performance of our method.

Weijie Huang, Zhiping Li

A mixed finite element method (MFEM), using dual-parametric piecewise bi-quadratic and affine (DP-Q2-P1) finite element approximations for the deformation and the pressure like Lagrange multiplier respectively, is developed and analyzed for the numerical computation of incompressible nonlinear elasticity problems with large deformation gradient, and a damped Newton method is applied to solve the resulted discrete problem. The method is proved to be locking free and stable. The accuracy and efficiency of the method are illustrated by numerical experiments on some typical cavitation problems.

Weijie Huang, Jun Wang

This article provides an interesting exploration of character-level convolutional neural network solving Chinese corpus text classification problem. We constructed a large-scale Chinese language dataset, and the result shows that character-level convolutional neural network works better on Chinese corpus than its corresponding pinyin format dataset. This is the first time that character-level convolutional neural network applied to text classification problem.