Min Huang

EAAI Journal 2026 Journal Article

A social recommendation model based on cross-view contrastive learning and multi-head attention for multi-rating fusion

• Rui Chen
• Zhuo Dai
• Wei Lu
• Yanbu Guo
• Weizhi Meng
• Pu Li
• Min Huang
• Xiangjie Kong

In recent years, social recommender systems have become a hot research field. Contrastive learning effectively enhances the expressiveness of user representations by modeling the consistency of representations between interactive views and social views, thereby improving recommendation performance. This paper proposes a social recommendation model based on cross-view contrastive learning, which employs a multi-head attention mechanism to fuse multi-rating information. It adaptively assigns weights to multiple views, making more effective use of rich social relationships and social trust information to alleviate data sparsity. In the rating view, interaction-aware noise with orientation-preserving constraints is introduced for data augmentation. The proposed model constructs a cross-view contrastive learning task between the rating view and the social view. Extensive experiments on three real-world datasets demonstrate the effectiveness of the proposed approach and its key components, and reveal that our model consistently outperforms state-of-the-art methods.

EAAI Journal 2026 Journal Article

A two-stage evidence fusion underwater target recognition method based on Multivariate Variational Mode Decomposition feature optimization and conflict handling

• Jiahui Dai
• Yibo Zou
• Zhongzhe Xiao
• Min Huang

Underwater acoustic target recognition plays a critical role in marine safety, environmental monitoring, and national security. However, the complex and dynamic nature of underwater environments — characterized by noise interference, signal attenuation, and multi-path propagation — poses significant challenges to reliable and accurate target identification. To address these issues, this paper proposes a novel, robust framework for underwater target recognition by integrating an enhanced signal decomposition technique with a conflict-aware evidence fusion strategy. We first introduce MVMD-CSFR (Multivariate Variational Mode Decomposition with Component-Specific Feature Refinement), which combines MVMD with grey relational analysis for adaptive mode decomposition and effective component selection. This step significantly improves signal clarity by eliminating noise-dominated and redundant modes. High-dimensional features are then extracted from the refined modes through time–frequency, spectral, auditory, and waveform analyses, followed by Fisher score-based feature selection to reduce redundancy and enhance discriminability. To further improve decision reliability in uncertain environments, we propose a Two-Stage Conflict-Aware Evidence Theory (TSCAET) fusion framework that dynamically identifies and corrects conflicting evidence from heterogeneous mode components. Evaluated on real-world underwater acoustic datasets, the proposed method achieves a recognition accuracy of 90. 64%, representing a 13. 07% improvement over baseline approaches. Ablation studies confirm the individual contributions of MVMD-CSFR (improvement: 0. 97%–8. 26%) and TSCAET (improvement: 1. 51%–10. 37%) across diverse conditions. This work advances the state-of-the-art in underwater acoustic recognition by offering a principled, data-driven pipeline that enhances both feature quality and decision fusion under uncertainty.

EAAI Journal 2026 Journal Article

Corrigendum to “A two-stage evidence fusion underwater target recognition method based on Multivariate Variational Mode Decomposition feature optimization and conflict handling” [Eng. Appl. Artif. Intell. 163, (Part 1), (1 January 2026), 112769]

• Jiahui Dai
• Yibo Zou
• Zhongzhe Xiao
• Min Huang

AAAI Conference 2026 Conference Paper

DanceHA: A Multi-Agent Framework for Document-Level Aspect-Based Sentiment Analysis

• Lei Wang
• Min Huang
• Eduard Dragut

Aspect-Based Sentiment Intensity Analysis (ABSIA) has garnered increasing attention, though research largely focuses on domain-specific, sentence-level settings. In contrast, document-level ABSIA--particularly in addressing complex tasks like extracting Aspect-Category-Opinion-Sentiment-Intensity (ACOSI) tuples--remains underexplored. In this work, we introduce DanceHA, a multi-agent framework designed for open-ended, document-level ABSIA with informal writing styles. DanceHA has two main components: Dance, which employs a divide-and-conquer strategy to decompose the long-context ABSIA task into smaller, manageable sub-tasks for collaboration among specialized agents; and HA, Human-AI collaboration for annotation. We release Inf-ABSIA, a multi-domain document-level ABSIA dataset featuring fine-grained and high-accuracy labels from DanceHA. Extensive experiments demonstrate the effectiveness of our agentic framework and show that the multi-agent knowledge in DanceHA can be effectively transferred into student models. Our results highlight the importance of the overlooked informal styles in ABSIA, as they often intensify opinions tied to specific aspects.

EAAI Journal 2025 Journal Article

Industrial device-aided data collection for real-time rail defect detection via a lightweight network

• Qing Dong
• Tianxin Han
• Gang Wu
• Lina Sun
• Min Huang
• Fu Zhang

Rail defect detection is challenging due to the diverse and irregular nature of defects, along with the limited availability of high-quality datasets. Existing methods struggle with effectively capturing multi-scale features for proper feature allocation and preserving crucial details in deep networks, leading to incomplete defect representation and reduced accuracy. To address these limitations, we propose Rail Defect Detection Network (REDNet), a lightweight deep learning model specifically designed for real-time rail defect detection in manufacturing and maintenance applications. We design the Multi-Scale Deep Feature Aggregation (MSDFA) module to enhance semantic consistency modeling and achieve more precise feature fusion. We develop the Adaptive Task Decomposition Head (ATDH) to address dynamic feature allocation, and we introduce the Reversible Column Network (RevCol) as the backbone to enhance feature extraction and ensure information reconstruction. Additionally, we developed a high-quality dataset using specialized equipment to address data scarcity and utilized it for training. REDNet achieved a high mean Average Precision at an Intersection over Union (IoU) threshold of 0. 50 (mAP50) of 94. 1% with exceptional real-time performance at 204. 1 frames per second (FPS), while keeping an efficient design of 5. 70 million parameters and surpassing state-of-the-art (SOTA) methods in accuracy and speed. These features make it suitable for defect detection, facilitating engineering deployment, and improve quality control in rail manufacturing and maintenance. Generalization tests on the public Microsoft Common Objects in Context (MS COCO) dataset yielded a mean Average Precision across IoU thresholds from 0. 50 to 0. 95 (mAP50–95) of 46. 8%, further confirming the effectiveness of REDNet.

EAAI Journal 2025 Journal Article

Underwater acoustic target detection with imbalanced dataset based on deep Q-network with dual termination criterion

• Zhifeng Liu
• Chenrui Hu
• Ben Xu
• Min Huang
• Zhongzhe Xiao

Underwater acoustic target detection often encounters the challenge of severe data imbalance. This work proposes an improved deep reinforcement learning algorithm for imbalanced classification, termed DQNiDTC(Deep Q-Network Based Algorithm for Imbalanced Classification with Dual Termination Condition). To fully utilize prior knowledge of acoustic signals, we propose an underwater acoustic feature acquisition strategy for imbalanced dataset with methods of feature extraction and feature selection, which helps to simplify the structure of Q-network from CNNs(Convolutional Neural Networks) to FCNNs(Fully Convolutional Networks) for faster convergence. To correctly classify samples of target echoes and radiated noises with serious imbalanced distribution, we propose a dual termination criterion for the simulation of designed environment, where misjudgments of both positive and negative samples will lead to the termination of that round of training with a reward mechanism for imbalanced classification, which enhances imbalanced classification by effectively utilizing rewards for both positive and negative samples, minimizing bias toward the majority class. Compared with existing method of DQNimb(Deep Q-Network Based Algorithm for Imbalanced Classification), experimental results show that F-measure of DQNiDTC is improved by 4. 59%–10. 23% with the imbalanced level of 1: 100, 1: 150, 1: 200 and 1: 250, thereby achieving efficient imbalanced binary classification for underwater acoustic target detection. Additionally, we provide the design and testing of a quasi-real-time model prediction based on the DQNiDTC algorithm, which can meet the quasi-real-time requirements of the model prediction end in imbalanced underwater acoustic target detection systems.

IJCAI Conference 2024 Conference Paper

Towards Robust Multi-Label Learning against Dirty Label Noise

• Yuhai Zhao
• Yejiang Wang
• Zhengkui Wang
• Wen Shan
• Miaomiao Huang
• Meixia Wang
• Min Huang
• Xingwei Wang

In multi-label learning, one of the major challenges is that the data are associated with label noise including the random noisy labels (e. g. , data encoding errors) and noisy labels created by annotators (e. g. , missing, extra, or error label), where noise is promoted by different structures (e. g. , gaussian, sparse or subjective). Existing methods are tailored to handle noise with one specific structure. However, they lack of consideration of the fact that the data are always with dirty noisy labels, simutaneously gaussian, sparse and subjective, in real applications. In this paper, we formalize the multi-label learning with dirty noise as a new learning problem, namely Noisy Multi-label Learning (NML). To solve the NML problem, we decompose a corrupted label matrix as the noise matrix plus a true label matrix (maybe high-rank). For the noise matrix, a mixed norm penalty is developed as regularizer for dirty noise distribution. Under this norm, the conditions required for exact noise recovery are provided theoretically. For the true label matrix that is not necessarily low-rank, we apply a non-linear mapping to ensure its low-rankness such that the high-order label correlation can be utilized. Experimental results show that the proposed method outperforms the state-of-the-art methods significantly.

EAAI Journal 2021 Journal Article

Winner determination for logistics service procurement auctions under disruption risks and quantity discounts

• Mingqiang Yin
• Xiaohu Qian
• Min Huang
• Qingyu Zhang

Disruption risks and quantity discounts have not been explicitly studied in traditional winner determination problems. This paper investigates an innovative winner determination problem with accidental disruptions and quantity discounts for a fourth party logistics provider in a logistics service procurement auction. An effective hybrid strategy that integrates a fortification policy with an outside option policy is developed to mitigate possible disruptions. Integrating the hybrid strategy with a piecewise-linear discount function, a stochastic nonlinear winner determination model is constructed. Utilizing a linearization technique, such a nonlinear winner determination model can be reformulated as a mixed integer linear program. To obtain nearly optimal solutions, a scenario-based approximation method is developed, including a reduced scenario technique and a dual decomposition Lagrangian relaxation technique. Numerical experiments are conducted to illustrate the effectiveness and applicability of the proposed model and method. When the discount parameter decreases or the disruption probability increases, the fortification policy becomes more important for the logistics system. An interesting result indicates that both the auctioneer and bidders may benefit from considering the discount policy. Managerial implications are drawn for fourth party logistics providers to run a reliable logistics system in a cost-effective way.

EAAI Journal 2020 Journal Article

Two-level principal–agent model for schedule risk control of IT outsourcing project based on genetic algorithm

• Hualing Bi
• Fuqiang Lu
• Shupeng Duan
• Min Huang
• Jinwen Zhu
• Mengying Liu

With increasing developments in the Information Technology (IT) outsourcing industry, many enterprises outsource IT services to reduce costs. However, the schedule risk of IT outsourcing (ITO) projects may result in enormous economic losses for an enterprise. In this paper, the principal–agent theory is used to control the schedule risk of ITO projects. A two-level mathematical model is built to describe the decision process of the client and vendors. With an increase to the number of subprojects and activities, the scale of the problem will become very large. The resulting optimization is an NP hard problem with continuous domain. Therefore, a genetic algorithm (GA) is designed to solve the proposed model. Experiments are performed to test the ability of the proposed algorithm. Some insights from simulation analysis – the principal–agent theory and two-level mathematical model – are suitable for describing the cooperative relationship between principle and agent. By comparing with ant colony optimization and simulated annealing, the proposed GA shows strong optimization abilities for convergence, reliability, and efficiency, which is a good tool for this kind of optimization problem. The near-optimal plan reduced the schedule risk of the project remarkably, which is the scientific quantitative proposal for the decision maker. This study provides practitioners insights on relationships of schedule risk and ITO projects, and the design model and algorithms of this paper provides practitioners effective potential method to reduce the schedule risk of ITO projects in their operations. However, the uncertain characteristics of key and multiple factors should be considered in future work. Stochastic Programming and the Monte Carlo Simulation Method are two potential tools for dealing with uncertain factors. Additionally, the proposed GA could potentially be improved in terms of convergence. The advantages of other intelligent algorithms could be applied to the GA in order to improve its searching ability, such as the Taboo mechanism.

EAAI Journal 2018 Journal Article

A systematic model of stable multilateral automated negotiation in e-market environment

• Taiguang Gao
• Min Huang
• Qing Wang
• Mingqiang Yin
• Wai Ki Ching
• Loo Hay Lee
• Xingwei Wang

In e-market environment, the participants are usually bilateral such as in Consumer-to-Business or Customer-to-Customer business models. The participant on each side prefers the counterpart from which the concerned issues or profits can be pursued. Hence, the effective matching from a global point of view and the stable matching from an individual point of view become the critical function of the business models. In this paper, a systematic model of Stable Multilateral Automated Negotiation (SMAN) is proposed to facilitate the involved parties’ matching process in two-sided e-market, where confidential mediator agent as well as party agents communicate and make decisions on behalf of their principal parties. To make the matching effective and stable, two optimization models are designed. One is matching points model which makes an effective balance among the proposal value of issues for each possible pair of matching, such that the joint weighted profit measure is optimized with feature rescaling. The other one is matching scheme model which optimizes Social Welfare (SW) subject to the stable constraints, and ensures the engaged individual party satisfies the matching result from its viewpoint. And the optimality of stable matching is proved by mathematical deduction. Finally, numerical experiments are illustrated and show that the designed systematic models can generate effective matchings with individually stable advantages over the traditional Multilateral Automated Negotiation of Two Sides (MANTS).

EAAI Journal 2013 Journal Article

Multi-robot navigation based QoS routing in self-organizing networks

• Xingwei Wang
• Hui Cheng
• Min Huang

The technical development drives the future networks to become large-scale, heterogeneous, and dynamic. Bio-inspired networking can help reduce the time-space complexity of the complex network. Due to the good features such as self-organization and self-management, self-organizing network (SON) will most probably be a priority choice for the next generation network. In this paper, a swarm intelligence based Quality of Service (QoS) routing protocol is proposed for SON. The inaccurate routing and QoS information is described with fuzzy mathematics whilst the utilities of both the user and the network service provider are considered by applying game theory. Based on the multi-robot navigation algorithm, the protocol is able to search a routing path which can satisfy the user QoS requirements and achieve the Pareto optimal utilities of the user and the network service provider under Nash equilibrium. The proposed protocol is implemented and evaluated by extensive simulation experiments. The results show that it beats both other swarm intelligence based routing protocols and the traditional Dijkstra algorithm based routing protocol. The searched routing paths support the win–win effect for both the user and the network service provider.