predictive neighbourhood selection

This study addresses the core challenge in decentralized railway traffic management systems: balancing local computational efficiency with global scheduling consistency in safety-critical scenarios. The authors propose a self-organizing coordination mechanism wherein trains generate scheduling hypotheses based on short-horizon predictions of their neighborhood and achieve agreement with neighboring trains through a distributed consensus algorithm. Notably, the work demonstrates that a surprisingly short prediction time window suffices to guarantee global scheduling feasibility, thereby significantly enhancing local computational efficiency and response speed—challenging the conventional wisdom that longer prediction horizons yield superior performance. The effectiveness of the proposed approach is validated through closed-loop simulations grounded in a dynamic social interaction graph model.

This study addresses a central question in network science: how ubiquitous global structural features of complex networks—such as hubs, short path lengths, and high clustering—emerge without access to global information. The work proposes that these macroscopic properties arise not from global mechanisms but through bottom-up emergence driven by simple local rules, wherein nodes connect solely based on information from their immediate neighbors. By constructing a purely local growth model, analyzing empirical networks across diverse domains—including citation, social, and protein–protein interaction networks—and providing intuitive theoretical explanations, the study demonstrates for the first time that a unified local rule can reproduce key topological characteristics of real-world networks without invoking global assumptions such as preferential attachment. This finding offers a new paradigm for understanding self-organization in complex systems.

We study agents playing a pure coordination game on a large social network. Agents are restricted to coordinate locally, without access to a global communication device, and so different regions of the network will converge to different actions, precluding perfect coordination. We show that the extent of this inefficiency depends on the network geometry: on some networks, near-perfect efficiency is achievable, while on others welfare is strictly bounded away from the optimum. We provide a geometric condition on the network structure that characterizes when near-efficiency is attainable. On networks in which it is unattainable, our results more generally preclude high correlations between outcomes in a large spectrum of dynamic games.

This study investigates the efficiency of neighborhood exploration strategies in multi-objective local search, with a focus on the performance gap between systematic traversal and random sampling. Through empirical analysis across diverse multi-objective optimization problems and supporting probabilistic modeling, the work provides the first theoretical and experimental evidence that random sampling consistently outperforms systematic exploration—including both best-improvement and first-improvement strategies. This advantage stems from the observation that high-quality neighboring solutions are sparsely and approximately uniformly distributed in the solution space, enabling random sampling to discover non-dominated solutions more efficiently at lower computational cost. These findings establish a new paradigm for designing multi-objective local search algorithms.

This work addresses three critical issues in evaluating Large Neighborhood Search (LNS) methods for Anytime Multi-Agent Path Finding (MAPF): inaccurate baseline performance, inconsistent evaluation metrics, and non-reproducible implementations of learning-based approaches. To this end, we introduce the first open-source, standardized LNS evaluation framework for MAPF. Through systematic reproduction and fair comparison of state-of-the-art LNS algorithms, we establish that rule-based heuristics—particularly CBS-LNS—serve as strong, robust baselines; existing supervised learning methods show no significant improvement in solution quality or runtime efficiency. Based on these findings, we propose three novel research directions: (1) handling high-latency agents, (2) context-aware re-planning, and (3) dynamic neighborhood size adaptation. Extensive experiments validate the robustness of rule-based baselines. We publicly release all code, trained models, and benchmark datasets to enable reproducible and comparable MAPF-LNS research.

本文通过图模型设计解决在有限理性下人机协作问题，利用图子模型和变分法优化网络拓扑结构以最大化全局协调度。

This study investigates how local social influence modulates individual decision-making in structured populations, where agents simultaneously weigh the intrinsic value of alternatives against the choices of their neighbors. To this end, we propose a theoretical model that integrates both sources into perceived utility, thereby formalizing— for the first time—a collective decision mechanism incorporating social influence within graph-structured groups. Methodologically, we combine analytical results from static weighted connected graphs with a Markov-switching framework for dynamic networks, validated through simulations. Our findings reveal that social influence can either amplify the advantage of high-quality options or compensate for the disadvantage of inferior ones. In dynamic networks, the collective outcome is jointly determined by the average degree of each network configuration and its expected dwell time, with theoretical predictions showing strong agreement with simulation results.

研究提出一种网络感知的治疗分配框架，通过优化Fisher信息矩阵来解决网络干扰下的实验设计问题，并开发了适用于大规模网络的高效局部搜索算法。

This study addresses the challenge that algorithm design for complex systems relies on costly Monte Carlo simulations, hindering efficient evaluation of long-term intervention effects. To overcome this, we propose employing action-conditioned world models as rapid evaluators within the algorithm design loop. Specifically, our method leverages diffusion dynamics learning to construct simulation environments encoding agent behaviors. Through full-trajectory rollback and counterfactual probing, it enables action-level credit assignment and surrogate intervention analysis, thereby facilitating iterative algorithm refinement. Experimental results demonstrate that the proposed approach outperforms baselines across eight tasks while achieving a 14.5× speedup in rollback compared to Monte Carlo simulations, significantly overcoming existing computational bottlenecks.

研究网络系统决策中信息范围与新鲜度之间的权衡，通过比较局部即时观察与全局延迟观察，提出最优信息架构。