Changjun Fan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Changjun Fan has a total of 105 co-authors. Here are the top co-authors by number of shared publications:
Zhong Liu
National University of Defense Technology
5 shared publications- Zhong Liu
Laboratory for Big Data and Decision, National University of Defense Technology, Changsha, China
5 shared publications - Jincai Huang
Hunan Institute of Advanced Technology, Changsha, China
4 shared publications Kuihua Huang
National University of Defense Technology
4 shared publications- Shixuan Liu
College of Computer Science and Technology, National University of Defense Technology, Changsha, China
4 shared publications Tianle Pu
National University of Defense Technology
4 shared publications
Latest Publications
- Open Access
- 21 September 2026
- Published by Institute of Central Computation and Knowledge Inc. in ICCK Transactions on Intelligent Systematics
- Vol. 3 (3) , 162-171
DECIS: An LLM-Based Value Assessment Framework for Targets
Target value assessment is a critical component of operational decision-making; however, existing methodologies suffer from the rigidity of static models in dynamic battlefield environments and the scarcity of specialized domain-specific data. To address these problems, we propose DECIS, a framework for adaptive decision support based on Retrieval-Augmented Generation (RAG). DECIS operates through a cohesive three-stage architecture centered on two core functions: first, RAG is employed to intelligently match optimal assessment strategies to dynamic contexts, thereby mitigating the challenge of limited military corpora; second, the Large Language Model (LLM) component autonomously generates and refines target lists, ensuring decision recommendations are characterized by high temporal validity and situational awareness. Experimental validation confirms that DECIS reduces the Largest Connected Component to 0.6, compared to 0.75 for PageRank and 0.8 for Betweenness Centrality, providing robust decision support and demonstrating competitive performance against established baselines in dynamic target value assessment.
- 21 September 2026
- Published by Institute of Central Computation and Knowledge Inc. in ICCK Transactions on Intelligent Systematics
- Vol. 3 (3) , 162-171
Open AccessDECIS: An LLM-Based Value Assessment Framework for Targets
Target value assessment is a critical component of operational decision-making; however, existing methodologies suffer from the rigidity of static models in dynamic battlefield environments and the scarcity of specialized domain-specific data. To address these problems, we propose DECIS, a framework for adaptive decision support based on Retrieval-Augmented Generation (RAG). DECIS operates through a cohesive three-stage architecture centered on two core functions: first, RAG is employed to intelligently match optimal assessment strategies to dynamic contexts, thereby mitigating the challenge of limited military corpora; second, the Large Language Model (LLM) component autonomously generates and refines target lists, ensuring decision recommendations are characterized by high temporal validity and situational awareness. Experimental validation confirms that DECIS reduces the Largest Connected Component to 0.6, compared to 0.75 for PageRank and 0.8 for Betweenness Centrality, providing robust decision support and demonstrating competitive performance against established baselines in dynamic target value assessment.
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 851-862
A Perturbation-Augmented Unsupervised Learning Framework for Integer Linear Programming
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 851-862
Open AccessA Perturbation-Augmented Unsupervised Learning Framework for Integer Linear Programming
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 556-567
Generating Graph-Like Logical Rules for Knowledge Graph Reasoning via Diffusion Models
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 556-567
Open AccessGenerating Graph-Like Logical Rules for Knowledge Graph Reasoning via Diffusion Models
- 2 July 2026
- Published by Springer Nature
- p. 45-53
Spin Glass Optimization: The Convergence of LLMs and Evolutionary Algorithms
- 2 July 2026
- Published by Springer Nature
- p. 45-53
Spin Glass Optimization: The Convergence of LLMs and Evolutionary Algorithms
- 2 July 2026
- Published by Springer Nature
- p. 66-74
DNKD: Self-Supervised Dual Network Knowledge Distillation for Constrained Optimization
- 2 July 2026
- Published by Springer Nature
- p. 66-74
DNKD: Self-Supervised Dual Network Knowledge Distillation for Constrained Optimization
- 2 July 2026
- Published by Springer Nature
- p. 21-31
Hyperbolic Give You More: Learning Combinatorial Optimization Algorithm over Graph in Hyperbolic Space
- 2 July 2026
- Published by Springer Nature
- p. 21-31
Hyperbolic Give You More: Learning Combinatorial Optimization Algorithm over Graph in Hyperbolic Space
- 12 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Engineering Management
- Vol. 73 (00189391) , 3723-3737
Identifying the Weak Set in Interdependent Networks Considering Cascading Failures
Identifying the weak set of interdependent networks that influences both structure and function is of great importance for studying infrastructure resilience and supporting engineering management. However, the vulnerability of interdependent networks arises not only from their structural properties but also from the cascading propagation of failures, which increases the complexity of weak set identification and makes it difficult for classical methods to balance both accuracy and efficiency. In this study, a data-driven framework, termed the GCN-WSI (Graph Convolutional Network-based Weak Set Identification model), is proposed to identify the weak set that can disintegrate interdependent networks under cascading failures. First, the cascading failure mechanism and related properties of interdependent networks are characterized, which reformulates weak set identification as a prediction problem of estimating the probability that nodes belong to the weak set. Second, the concept of dual-stream graph embedding is introduced to aggregate neighbor information within and across layers, incorporating attributes associated with structural properties and cascading failures to construct comprehensive representations. Finally, a multilayer perceptron (MLP) regressor is employed to estimate the probability of node membership in the weak set. The GCN-WSI framework is trained on small-scale synthetic networks and is applicable to large-scale real-world systems. Experiments on diverse datasets validate the effectiveness of the proposed method and highlight the necessity and feasibility of specific components. In particular, under predefined disintegration targets, the model demonstrates strong capability in identifying a compact weak set whose removal leads to a rapid decline in network performance.
- 12 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Engineering Management
- Vol. 73 (00189391) , 3723-3737
Identifying the Weak Set in Interdependent Networks Considering Cascading Failures
Identifying the weak set of interdependent networks that influences both structure and function is of great importance for studying infrastructure resilience and supporting engineering management. However, the vulnerability of interdependent networks arises not only from their structural properties but also from the cascading propagation of failures, which increases the complexity of weak set identification and makes it difficult for classical methods to balance both accuracy and efficiency. In this study, a data-driven framework, termed the GCN-WSI (Graph Convolutional Network-based Weak Set Identification model), is proposed to identify the weak set that can disintegrate interdependent networks under cascading failures. First, the cascading failure mechanism and related properties of interdependent networks are characterized, which reformulates weak set identification as a prediction problem of estimating the probability that nodes belong to the weak set. Second, the concept of dual-stream graph embedding is introduced to aggregate neighbor information within and across layers, incorporating attributes associated with structural properties and cascading failures to construct comprehensive representations. Finally, a multilayer perceptron (MLP) regressor is employed to estimate the probability of node membership in the weak set. The GCN-WSI framework is trained on small-scale synthetic networks and is applicable to large-scale real-world systems. Experiments on diverse datasets validate the effectiveness of the proposed method and highlight the necessity and feasibility of specific components. In particular, under predefined disintegration targets, the model demonstrates strong capability in identifying a compact weak set whose removal leads to a rapid decline in network performance.
- Open Access
Optimizing p-spin models through hypergraph neural networks and deep reinforcement learning
p-spin glasses, characterized by frustrated many-body interactions beyond the conventional pairwise case (p>2), are prototypical disordered systems whose ground-state search is NP-hard and computationally prohibitive for large instances. Solving this problem is not only fundamental for understanding high-order disorder, structural glasses, and topological phases, but also central to a wide spectrum of hard combinatorial optimization tasks. Despite decades of progress, there still lacks an efficient and scalable solver for generic large-scale p-spin models. Here we introduce PLANCK, a physics-inspired deep reinforcement learning framework built on hypergraph neural networks. PLANCK directly optimizes arbitrary high-order interactions, and systematically exploits gauge symmetry throughout both training and inference. Trained exclusively on small synthetic instances, PLANCK exhibits strong zero-shot generalization to systems orders of magnitude larger, and consistently outperforms state-of-the-art thermal annealing methods across all tested structural topologies and coupling distributions. Moreover, without any modification, PLANCK achieves near-optimal solutions for a broad class of NP-hard combinatorial problems, including random k-XORSAT, hypergraph max-cut, and conventional max-cut. The presented framework provides a physics-inspired algorithmic paradigm that bridges statistical mechanics and reinforcement learning. The symmetry-aware design not only advances the tractable frontiers of high-order disordered systems, but also opens a promising avenue for machine-learning-based solvers to tackle previously intractable combinatorial optimization challenges.
Open AccessOptimizing p-spin models through hypergraph neural networks and deep reinforcement learning
p-spin glasses, characterized by frustrated many-body interactions beyond the conventional pairwise case (p>2), are prototypical disordered systems whose ground-state search is NP-hard and computationally prohibitive for large instances. Solving this problem is not only fundamental for understanding high-order disorder, structural glasses, and topological phases, but also central to a wide spectrum of hard combinatorial optimization tasks. Despite decades of progress, there still lacks an efficient and scalable solver for generic large-scale p-spin models. Here we introduce PLANCK, a physics-inspired deep reinforcement learning framework built on hypergraph neural networks. PLANCK directly optimizes arbitrary high-order interactions, and systematically exploits gauge symmetry throughout both training and inference. Trained exclusively on small synthetic instances, PLANCK exhibits strong zero-shot generalization to systems orders of magnitude larger, and consistently outperforms state-of-the-art thermal annealing methods across all tested structural topologies and coupling distributions. Moreover, without any modification, PLANCK achieves near-optimal solutions for a broad class of NP-hard combinatorial problems, including random k-XORSAT, hypergraph max-cut, and conventional max-cut. The presented framework provides a physics-inspired algorithmic paradigm that bridges statistical mechanics and reinforcement learning. The symmetry-aware design not only advances the tractable frontiers of high-order disordered systems, but also opens a promising avenue for machine-learning-based solvers to tackle previously intractable combinatorial optimization challenges.
- 28 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Cognitive Communications and Networking
- Vol. 12, 5913-5936
Cognitive Underwater Acoustic Networking and Target Tracking: A Comprehensive Survey
Underwater acoustic networks are evolving from static, manually-configured systems into cognitive, learning-enabled platforms that can perceive, reason, and adapt to harsh ocean dynamics in real-time. Accurate target tracking is a core service of these networks and underpins marine resource exploration, environmental monitoring, and maritime security. Existing reviews or surveys, however, rarely examine underwater acoustic target tracking through the lens of cognitive communications and networking, and often offer a narrow perspective on addressing the paradigm shifts driven by emerging technologies like deep learning. To fill this gap, this work presents a systematic survey of this field and introduces an innovative three-dimensional taxonomy framework based on the three levels of the cognitive underwater acoustic target tracking network: the target layer, the perception layer, and the processing layer.Within this framework, we comprehensively survey the literature over the period 2016-2025, spanning from the theoretical foundations to diverse algorithmic approaches in underwater acoustic target tracking. Particularly, we emphasize the transformative potential and recent advancements of machine learning techniques, including deep learning and reinforcement learning, in enhancing the performance and adaptability of cognitive underwater tracking systems. Finally, this survey concludes by identifying key challenges in the field and proposing future avenues based on emerging technologies such as data desensitization, embodied intelligence, and large models.
- 28 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Cognitive Communications and Networking
- Vol. 12, 5913-5936
Cognitive Underwater Acoustic Networking and Target Tracking: A Comprehensive Survey
Underwater acoustic networks are evolving from static, manually-configured systems into cognitive, learning-enabled platforms that can perceive, reason, and adapt to harsh ocean dynamics in real-time. Accurate target tracking is a core service of these networks and underpins marine resource exploration, environmental monitoring, and maritime security. Existing reviews or surveys, however, rarely examine underwater acoustic target tracking through the lens of cognitive communications and networking, and often offer a narrow perspective on addressing the paradigm shifts driven by emerging technologies like deep learning. To fill this gap, this work presents a systematic survey of this field and introduces an innovative three-dimensional taxonomy framework based on the three levels of the cognitive underwater acoustic target tracking network: the target layer, the perception layer, and the processing layer.Within this framework, we comprehensively survey the literature over the period 2016-2025, spanning from the theoretical foundations to diverse algorithmic approaches in underwater acoustic target tracking. Particularly, we emphasize the transformative potential and recent advancements of machine learning techniques, including deep learning and reinforcement learning, in enhancing the performance and adaptability of cognitive underwater tracking systems. Finally, this survey concludes by identifying key challenges in the field and proposing future avenues based on emerging technologies such as data desensitization, embodied intelligence, and large models.
- 14 January 2026
- Published by Elsevier in Information Processing & Management
- Vol. 63 (4)
NeuPath: A hybrid learning-based optimization approach for emergency search path planning
- 14 January 2026
- Published by Elsevier in Information Processing & Management
- Vol. 63 (4)
NeuPath: A hybrid learning-based optimization approach for emergency search path planning