Chunguo Wu

Discovering governing Partial Differential Equations (PDEs) from observational data is a fundamental challenge in AI for Science. While Symbolic Regression (SR) dominates this task, existing token-level methods trigger a combinatorial explosion of search spaces, frequently yielding mathematically valid yet physically inconsistent candidates. To overcome this, we propose Modular Symbolic Regression with Physics Priors (MSR-PP), a knowledge-guided framework that conceptualizes one-dimensional PDEs as structured compositions of semantic modules (e.g., convection, diffusion) rather than random symbol sequences. MSR-PP employs an RL-optimized LSTM agent to sequentially construct equation skeletons based on syntactic and semantic dependencies, utilizing a modular search strategy to significantly prune the search space. Addressing the critical scarcity of standardized, high-fidelity benchmarks for 1D PDEs, we constructed an extended dataset for rigorous evaluation. Extensive experiments demonstrate that MSR-PP outperforms state-of-the-art baselines. Real-world validation on the NGSIM US101 traffic dataset shows MSR-PP successfully identifies a convection-diffusion model featuring a critical second-order term. Notably, the negative coefficient of this term (-u_xx) aligns with theoretical models, accurately capturing the string instability and wave-amplifying dynamics characteristic of stop-and-go traffic. Moreover, the discovered equation exhibits robust out-of-distribution (OOD) generalization on the unseen NGSIM I-80 dataset, underscoring MSR-PP's capability to uncover physically meaningful laws. Datasets: https://github.com/jinyangdu5/MSR-PP.

Prediction-based dynamic constrained multiobjective evolutionary algorithms (DCMOEAs) have received increasing attention for their ability to leverage historical information to provide high-quality initial populations for new environments. However, these algorithms suffer from deviation from the target constrained Pareto front (CPF) after dynamic response, due to prediction bias when the learned trend mismatches the actual environmental change. This deviation (referred to as response error) persists into the static optimization, degrading both convergence and coverage under a limited computational budget. To address this issue, this paper proposes a prediction-based DCMOEA based on temporal--spatial response error suppression. Along the temporal dimension, a feedback-corrected migration strategy with Kalman filtering is designed to compensate for the prediction bias accumulated over the environmental change sequence, suppressing the response error at its source. A theoretical analysis establishes the condition under which the feedback correction is guaranteed to reduce the centroid prediction error. Along the spatial dimension, a subspace anchor-based environmental selection strategy preserves at least one representative solution in each region of the objective space, preventing environmental selection from eliminating solutions in regions affected by the response error. Experimental results on two benchmark suites demonstrate that the proposed algorithm significantly outperforms seven state-of-the-art DCMOEAs in finding Pareto optimal solutions on the majority of instances. The results also demonstrate that addressing the response error is a key factor for improving the performance of prediction-based DCMOEAs, offering a new perspective for designing prediction-based DCMOEAs.

Point cloud video modeling not only has to address the natural irregularity of point clouds, but also the challenge of capturing spatial and temporal representation simultaneously. Current methods attempt to approximate the temporal dimension using several 3-D point cloud frame sequences but struggle in sparser conditions. Accurate point trajectory tracking is crucial for effectively capturing temporal dynamics, as point positions across different frames are often inconsistent, especially during rapid motion or at low frame rates. Conventional point tube operations aggregate motion features over fixed time windows but fail to capture rapidly changing scenes. Implicit tracking techniques are limited by quadratic time complexity, which restricts their practical use. In this article, we propose a native 4-D framework (N4DF) that guides the network to learn spatio-temporal dynamics from a native 4-D perspective. Furthermore, we devise a dynamic point spatio-temporal (DPST) convolution to adaptively select the optimal point-tracking strategy, which constructs local plane regions in anchor frames and propagates them to neighboring frames to evaluate point cross-frame movement distances. To further enhance the global modeling power of N4DF, we develop a dynamic self-tracking re-encoding (DSTR) module that employs point-wise self-attention to search for relevant points across the entire video. Compared with the recent 4-D modeling methods, N4DF demonstrates superior performance on MSR-Action3D and NTU RGB+D for action recognition (+0.7% and +1.2% accuracy, respectively), on HOI4D for action segmentation (+1% accuracy), and on Synthia 4-D and nuScenes-lidarseg for semantic segmentation (+0.49% and +1.7% mIoU, respectively). Our N4DF shows greater robustness at low frame-rate settings due to native 4-D modeling and adaptive tracking, making it suitable for tracking fast-moving objects in future real-time scenarios.

Distributed Flexible Job Shop Scheduling with Transportation Constraints (DFJSP-T) is widely used in manufacturing workshops across various industries, as it more accurately represents real-world production scenarios. However, many existing studies do not fully leverage critical problem features or domain knowledge, limiting the ability of current algorithms to find high-quality solutions for complex large-scale problems. To address these challenges, this paper proposes a Self-Adaptive Differential Evolution Enhanced Quality-Diversity Optimization (SADE-QD). The SADE-QD incorporates domain knowledge in three main ways: (1) it models the problem using machine idle and job transportation features, allowing the algorithm to retain solutions with diverse behaviors; (2) it introduces a knowledge-guided heuristic search strategy based on critical path to discover more high-quality solutions; and (3) it applies an self-adaptive differential evolution search method that leverages machine idle and transportation features to explore a broader range of solutions. This helps valuable information from the feature space contribute directly to the search process. Experiments on small, medium, and large scale benchmarks show that SADE-QD achieves an average makespan reduction of 12% compared to several recent state-of-the-art algorithms.