Author name cluster

Xin Peng

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

10 papers

1 author row

AAAI Conference 2026 Conference Paper

Cross-View Progressive Feature Filtering for Multi-View Graph Clustering in Remote Sensing

Bowen Liu
Xin Peng
Wenxuan Tu
Chengyao Wei
Xiangyan Tang
Jieren Cheng
Miao Yu

Multi-view clustering of remote sensing data plays a vital role in Earth observation analysis. Recently, deep graph clustering methods based on contrastive learning have significantly improved feature representation capabilities. However, most existing approaches treat all views equally, neglecting the inherent uniqueness and heterogeneity across views, which often results in two major issues: 1) discriminative features from clustering-friendly views are underexplored; and 2) redundant or noisy information from less informative views can degrade the shared representation. To address these challenges, we propose a novel multi-view graph clustering framework termed CF-MVGC for remote sensing data, which dynamically preserves discriminative features and suppresses redundancy by assessing view affinity. Specifically, we employ a dual-stage representation learning strategy to extract both view-specific discriminative features and cross-view consistent representations. To further exploit and adaptively integrate complementary information across views, we design a progressive feature filtering model that dynamically evaluates view affinity using two novel metrics, i.e., view fidelity index (VFI) and view criticality index (VCI). Based on these assessments, the module adaptively modulates feature update and reset signals, reinforcing informative views while suppressing noisy or redundant ones. Views with high affinity receive strengthened update signals to retain valuable features, while those with low affinity are subjected to enhanced reset operations to eliminate noise and redundancy. The resulting high-quality, discriminative representations lead to improved clustering performance, establishing a positive feedback loop. Experimental results on four benchmark datasets demonstrate the effectiveness and superiority of CF-MVGC against its competitors.

PDF Details DOI

AAAI Conference 2026 Conference Paper

ExpertAD: Enhancing Autonomous Driving Systems with Mixture of Experts

Haowen Jiang
Xinyu Huang
You Lu
Dingji Wang
Yuheng Cao
Chaofeng Sha
Bihuan Chen
Keyu Chen

Recent advancements in end-to-end autonomous driving systems (ADSs) underscore their potential for perception and planning capabilities. However, challenges remain. Complex driving scenarios contain rich semantic information, yet ambiguous or noisy semantics can compromise decision reliability, while interference between multiple driving tasks may hinder optimal planning. Furthermore, prolonged inference latency slows decision-making, increasing the risk of unsafe driving behaviors. To address these challenges, we propose ExpertAD, a novel framework that enhances the performance of ADS with Mixture of Experts (MoE) architecture. We introduce a Perception Adapter (PA) to amplify task-critical features, ensuring contextually relevant scene understanding, and a Mixture of Sparse Experts (MoSE) to minimize task interference during prediction, allowing for effective and efficient planning. Our experiments show that ExpertAD reduces average collision rates by up to 20% and inference latency by 25% compared to prior methods. We further evaluate its multi-skill planning capabilities in rare scenarios (e.g., accidents, yielding to emergency vehicles) and demonstrate strong generalization to unseen urban environments. Additionally, we present a case study that illustrates its decision-making process in complex driving scenarios.

PDF Details DOI

EAAI Journal 2026 Journal Article

Large-scale stochastic production decision-making for coupled economy-environment-energy systems in sustainable industrial processes under uncertainty: A data-driven two-stage multi-objective optimization framework

Tingwei Zhang
Weimin Zhong
Shuai Tan
Feifei Shen
Yurong Liu
Xin Peng

Large-scale complex industrial processes feature highly coupled production environments and complex dynamic unit operations, hindering the low-carbon decision optimization under uncertainty and ultimately limiting production schemes’ flexibility and reliability. Accordingly, this paper develops a data-driven two-stage stochastic optimization framework considering the economic-environment-energy coupling effects to support sustainable production decision-making under uncertainty. Specifically, a machine learning-assisted economic-environment-energy coupling assessment method is proposed to accurately track economic, environmental, and energy footprints. Subsequently, a large-scale multi-objective production optimization model based on the economic-environment-energy assessment mechanism is established, incorporating a carbon tax scenario to efficiently obtain optimal low-carbon production schemes. Additionally, given that existing production optimization methods generally neglect the impact of carbon tax uncertainty, a two-stage stochastic programming framework is developed to enhance the flexibility of the production systems. The first stage implements a deterministic multi-objective optimization model to obtain the baseline production schemes. The second stage builds upon this and employs data-driven techniques for intelligent identification of carbon tax scenarios, deriving their probability distributions to optimize production decisions under uncertainty, thereby enabling dynamic adjustments to production structures and material flows. A case study applying the proposed model and methods to an actual integrated refinery-petrochemical production site demonstrates that the proposed approach can identify plant-wide carbon reduction opportunities while significantly improving economic performance. Specifically, the approach leads to an 8. 5% reduction in production costs and a 7% improvement in emissions assessment accuracy. The proposed method exhibits notable superiority in improving the model’s adaptability to policy fluctuations and in reinforcing the flexibility of the optimization schemes, which collectively provide reliable decision support for efficient resource utilization and sustainable development of industrial processes under uncertain environments.

Details DOI

EAAI Journal 2025 Journal Article

An efficient and lightweight adaptive network for three-dimensional medical image segmentation

Dayu Tan
Manman Shi
Yansen Su
Xin Peng
Chunhou Zheng
Kaixun He
Weimin Zhong

The Vision Transformer (ViT) demonstrates exceptional performance in three-dimensional medical image segmentation by introducing a non-local self-attention mechanism, thereby providing a larger receptive field. However, the increase in model capacity and dataset scale, along with the lack of image-specific biases and scaling behaviors, affects the accuracy and detail of segmentation results, particularly when dealing with large-scale three-dimensional images that require capturing local information. In this study, we re-evaluate and improve the architecture of Convolutional Neural Networks (CNNs) by constructing a lightweight three-dimensional convolutional network (ConvNet++). We revisit volumetric convolution through the concept of large-kernel deep convolution modules and implement independent linear depth convolution scaling for each channel feature. This improvement not only contributes to the architecture of deep learning models in the field of artificial intelligence but also provides a more efficient solution for medical image segmentation. We design a feature fusion method to better utilize multi-scale spatial information, avoiding simple addition or concatenation, thereby alleviating issues related to excessive model parameters and ineffective feature fusion. Additionally, we introduce a Lightweight Hybrid Attention (LHA) module that employs a self-attention mechanism with a shared key-query scheme to efficiently encode spatial and channel information. The LHA module facilitates effective communication between spatial and channel branches, providing complementary features while reducing the overall number of parameters. This design represents an innovation in adaptive modules within the field of artificial intelligence, enabling better capture of subtle features in medical images. We conduct extensive evaluations on four challenging benchmark datasets, and the results indicate that our contributions are highly effective in terms of parameter efficiency and accuracy. In the Brain tumor segmentation benchmark test, our model achieves a Dice coefficient of 85. 37%, with a model complexity of only 27. 39 million parameters, highlighting its lightweight and efficient characteristics. A comprehensive comparison of the model’s inference time is conducted on the Fluorodeoxyglucose Positron Emission Tomography/Computed Tomography dataset. The results indicate that 3D ConvNet++ performs relatively well in inference efficiency, with our model’s speed showing an approximately 3-fold improvement compared to 3D UX-NET and MedNeXt-B-K5. This achievement not only showcases the potential of artificial intelligence in the field of medical image segmentation but also provides an efficient solution for practical applications in medical engineering, assisting physicians in more accurately diagnosing diseases and planning treatments. Code: https: //github. com/6018203135/ConvNet.

Details DOI

EAAI Journal 2025 Journal Article

Causal explanation of nitrogen oxide emission predictions for fluid catalytic cracking unit based on convergent cross mapping: Predict the future and explain how

Han Jiang
Shucai Zhang
Jingru Liu
Xin Peng
Weimin Zhong

Industrial chemical processes are inherently intricate, characterized by prolonged operational sequences and correlations among features. Solely utilizing temporal information limits the prediction precision of deep learning methods. Moreover, causative features identified based on data may not align with the principles of chemical process. To solve these problems, this study proposes a spatial–temporal information based deep learning method, attention-based temporal graph convolutional network and convergent cross mapping (ATGCN-CCM). The devices are abstracted as nodes in a computational graph (CG) to represent the process, enabling the incorporation of spatial information into the predictive model. Attention mechanism is conducted within each node to dynamically weight the features. The CG is also used to select input features for causal analysis, ensuring that the identified causative features are not only consistent with the characteristics of the data, but also with the prior knowledge of the process. ATGCN-CCM is applied to datasets from industrial fluid catalytic cracking (FCC) units for nitrogen oxides (NOx) concentration prediction and causative feature identification. The prediction results demonstrate superior precision of ATGCN-CCM compared to some state-of-the-art spatial feature based, temporal feature based and hybrid methods. The identified features exhibit strong alignment with the principles of the chemical processes and the field experiences, thereby significantly enhancing model interpretability. The proposed ATGCN-CCM method illustrate its advanced capabilities in both precision and robustness, compared with attention-based methods and other causal analysis methods.

Details DOI

AAAI Conference 2025 Conference Paper

Federated Graph-Level Clustering Network

Jingxin Liu
Jieren Cheng
Renda Han
Wenxuan Tu
Jiaxin Wang
Xin Peng

Federated graph learning (FGL), which excels in analyzing non-IID graphs as well as protecting data privacy, has recently emerged as a hot topic. Existing FGL methods usually train the client model using labeled data and then collaboratively learn a global model without sharing their local graph data. However, in real-world scenarios, the lack of data annotations impedes the negotiation of multi-source information at the server, leading to sub-optimal feedback to the clients. To address this issue, we propose a novel unsupervised learning framework called Federated Graph-level Clustering Network (FedGCN), which collects the topology-oriented features of non-IID graphs from clients to generate global consensus representations through multi-source clustering structure sharing. Specifically, in the client, we first preserve the prototype features of each cluster from the structure-oriented embedding through clustering and then upload the learned multiple prototypes that are hard to be reconstructed into the raw graph data. In the server, we generate consensus prototypes from multiple condensed structure-oriented signals through Gaussian estimation, which are subsequently transferred to each client to promote the great encoding capacity of the local model for better clustering. Extensive experiments across multiple non-IID graph datasets have demonstrated the effectiveness and superiority of FedGCN against its competitors.

PDF Details DOI

EAAI Journal 2024 Journal Article

A truncated Gaussian distribution based multi-scale segment-wise fusion transformer model for multi-step commodity price forecasting

Xin Peng
Zhengxiang Chen
Jiale Zhang
Zhi Li
Wenli Du

Accurately forecasting commodity price trends is crucial for producers, market participants, and related enterprises to make informed decisions regarding production planning and scheduling. However, achieving high accuracy in multi-step forecasting poses significant challenges due to the unique financial characteristics inherent in commodities. Thus, this paper proposes a novel truncated Gaussian distribution based multi-scale segment-wise fusion Transformer for multi-step commodity price forecasting. First, a multi-scale segment-wise fusion module, which capture the time dependencies from different time granularity, is designed to describe the time-varying trend characteristics of commodity prices. Second, considering the characteristics of price range fluctuation and truncation, a truncated Gaussian distribution is introduced to describe price uncertainty. Last, to evaluate the proposed method’s effectiveness, extensive experiments are conducted using real data on energy chemical product prices. The experimental results demonstrate that the proposed method accurately captures price change trends and effectively estimates price uncertainty. Compared to the widely adopted Autoformer, our approach achieves approximately 30% reductions in both root mean square error (RMSE) and mean absolute error (MAE) metrics. Additionally, it exhibits certain advantages over the current state-of-the-art (SOTA). In the 20-step and 60-step multi-step prediction tasks, the proposed method achieves RMSE values of 91. 18 and 142. 94, respectively, surpassing the current SOTA. The introduced research framework provides valuable insights for decision-makers engaged in analyzing and forecasting commodity markets. The code is available on https: //github. com/dean-ob/TGD-MSSF.

Details DOI

AAAI Conference 2024 Conference Paper

Deep clustering with attribute-missing graphs, where only a subset of nodes possesses complete attributes while those of others are missing, is an important yet challenging topic in various practical applications. It has become a prevalent learning paradigm in existing studies to perform data imputation first and subsequently conduct clustering using the imputed information. However, these ``two-stage" methods disconnect the clustering and imputation processes, preventing the model from effectively learning clustering-friendly graph embedding. Furthermore, they are not tailored for clustering tasks, leading to inferior clustering results. To solve these issues, we propose a novel Attribute-Missing Graph Clustering (AMGC) method to alternately promote clustering and imputation in a unified framework, where we iteratively produce the clustering-enhanced nearest neighbor information to conduct the data imputation process and utilize the imputed information to implicitly refine the clustering distribution through model optimization. Specifically, in the imputation step, we take the learned clustering information as imputation prompts to help each attribute-missing sample gather highly correlated features within its clusters for data completion, such that the intra-class compactness can be improved. Moreover, to support reliable clustering, we maximize inter-class separability by conducting cost-efficient dual non-contrastive learning over the imputed latent features, which in turn promotes greater graph encoding capability for clustering sub-network. Extensive experiments on five datasets have verified the superiority of AMGC against competitors.

PDF Details DOI

IJCAI Conference 2024 Conference Paper

Synthesizing Programmatic Policy for Generalization within Task Domain

Tianyi Wu
Liwei Shen
Zhen Dong
Xin Peng
Wenyun Zhao

Deep reinforcement learning struggles to generalize across tasks that remain unseen during training. Consider a neural process observed in humans and animals, where they not only learn new solutions but also deduce shared subroutines. These subroutines can be applied to tasks involving similar states to improve efficiency. Inspired by this phenomenon, we consider synthesizing a programmatic policy characterized by a conditional branch structure, which is capable of capturing subroutines and state patterns. This enables the learned policy to generalize to unseen tasks. The architecture of the programmatic policy is synthesized based on a context-free grammar. Such a grammar supports a nested If-Then-Else derivation and the incorporation of Recurrent Neural Network. The programmatic policy is trained across tasks in a domain through a meta-learning algorithm. We evaluate our approach in benchmarks, adapted from PDDLGym for task planning and Pybullet for robotic manipulation. Experimental results showcase the effectiveness of our approach across diverse benchmarks. Moreover, the learned policy demonstrates the ability to generalize to tasks that were not seen during training.

PDF Details DOI

NeurIPS Conference 2023 Conference Paper

Enhancing Robot Program Synthesis Through Environmental Context

Tianyi Chen
Qidi Wang
Zhen Dong
Liwei Shen
Xin Peng

Program synthesis aims to automatically generate an executable program that conforms to the given specification. Recent advancements have demonstrated that deep neural methodologies and large-scale pretrained language models are highly proficient in capturing program semantics. For robot programming, prior works have facilitated program synthesis by incorporating global environments. However, the assumption of acquiring a comprehensive understanding of the entire environment is often excessively challenging to achieve. In this work, we present a framework that learns to synthesize a program by rectifying potentially erroneous code segments, with the aid of partially observed environments. To tackle the issue of inadequate attention to partial observations, we propose to first learn an environment embedding space that can implicitly evaluate the impacts of each program token based on the precondition. Furthermore, by employing a graph structure, the model can aggregate both environmental and syntactic information flow and furnish smooth program rectification guidance. Extensive experimental evaluations and ablation studies on the partially observed VizDoom domain authenticate that our method offers superior generalization capability across various tasks and greater robustness when encountering noises.

PDF Details