arXiv subjects

Huangyue Chen

Publications and source records attributed to Huangyue Chen.

4 recordsLinked to original sources

arXiv2026

OptGraph: Large Language Models Enhanced Evolutionary Optimization Via Graph Retrieval-Augmented Generation

Large language models (LLMs) have emerged as a powerful tool for automated evolutionary optimization, but existing methods remain limited in pattern reuse, error-aware refinement, and retrieval robustness across diverse tasks. To address these limitations, we propose OptGraph, the first optimization agentic workflow that introduces graph retrieval-augmented generation (GraphRAG). Specifically, OptGraph first constructs reusable experience as a typed graph, capturing the relationships among modeling patterns, problem formalization, implementation details, and error corrections. In the inference stage, OptGraph leverages graph neighborhood information to enrich retrieved knowledge, providing structured context to improve modeling, verification, and iterative refinement. Moreover, OptGraph supports adaptive knowledge updates, enabling the distillation of execution traces and verification feedback into reusable graph knowledge without ndertaking LLM parameter tuning. Extensive experiments on benchmark datasets show that our proposed OptGraph achieves an average exact accuracy 8.9% higher than the state-of-the-art prompt-based automated optimization frameworks. Our code has been made available at https://github.com/xianchaoxiu/OptGraph.

math.OC↗

arXiv2026

A Riemannian conjugate subgradient method for nonconvex and nonsmooth optimization on manifolds

Conjugate gradient (CG) methods are widely acknowledged as efficient for minimizing continuously differentiable functions in Euclidean spaces. In recent years, various CG methods have been extended to Riemannian manifold optimization, but existing Riemannian CG methods are confined to smooth objective functions and cannot handle nonsmooth ones. This paper proposes a Riemannian conjugate subgradient method for a class of nonconvex, nonsmooth optimization problems on manifolds. Specifically, we first select a Riemannian subgradient from the convex hull of two directionally active subgradients. The search direction is then defined as a convex combination of the negative of this subgradient and the previous search direction transported to the current tangent space. Additionally, a Riemannian line search with an interval reduction procedure is integrated to generate an appropriate step size, ensuring the objective function values form a monotonically nonincreasing sequence. We establish the global convergence of the algorithm under mild assumptions. Numerical experiments on three classes of Riemannian optimization problems show that the proposed method takes significantly less computational time than related existing methods. To our knowledge, this is the first CG-type method developed for Riemannian nonsmooth optimization.

math.OC↗

arXiv2025

Multi-View Clustering Meets Heterogenous Data: A Fusion Regularized Method

Multi-view clustering leverages consistent and complementary information across multiple views to provide more comprehensive insights than single-view analysis. However, the heterogeneity and redundancy of multi-view data pose significant challenges to the existing clustering techniques. To tackle these challenges effectively, this paper proposes a novel multi-view fusion regularized clustering method with adaptive group sparsity, enabling discriminative clustering while capturing informative features. Technically, for heterogeneous multi-view data with mixed-type feature sets, different losses or divergence metrics are considered with a joint fusion penalty to obtain consistent cluster structures. Moreover, the non-convex group sparsity consisting of inter-group sparsity and intra-group sparsity is utilized to eliminate redundant features, thereby enhancing the robustness. Furthermore, we develop an effective alternating direction method of multipliers (ADMM), where all subproblems can be solved in closed form. Extensive numerical experiments on real data validate the superior performance of our presented method in clustering accuracy and feature selection.

math.OC↗

arXiv2024

Safe Feature Identification Rule for Fused Lasso by An Extra Dual Variable

Fused Lasso was proposed to characterize the sparsity of the coefficients and the sparsity of their successive differences for the linear regression. Due to its wide applications, there are many existing algorithms to solve fused Lasso. However, the computation of this model is time-consuming in high-dimensional data sets. To accelerate the calculation of fused Lasso in high-dimension data sets, we build up the safe feature identification rule by introducing an extra dual variable. With a low computational cost, this rule can eliminate inactive features with zero coefficients and identify adjacent features with same coefficients in the solution. To the best of our knowledge, existing screening rules can not be applied to speed up the computation of fused Lasso and our work is the first one to deal with this problem. To emphasize our rule is a unique result that is capable of identifying adjacent features with same coefficients, we name the result as the safe feature identification rule. Numerical experiments on simulation and real data illustrate the efficiency of the rule, which means this rule can reduce the computational time of fused Lasso. In addition, our rule can be embedded into any efficient algorithm and speed up the computational process of fused Lasso.

stat.CO↗