Yao Lu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Yao Lu has a total of 46 co-authors. Here are the top co-authors by number of shared publications:
- Qi Xuan
Zhejiang University of Technology
10 shared publications - Yun Lin
Harbin Engineering University
9 shared publications - Dongwei Xu
College of Geoinformatics (College of Spatio-Temporal Intelligence Engineering), Zhejiang University of Technology, Hangzhou 313299, China
8 shared publications - Guan Gui
Nanjing University of Posts and Telecommunications
6 shared publications - Zeyu Wang
Institute of Cyberspace Security, College of Information Engineering, Zhejiang University of Technology, Hangzhou, China
3 shared publications Qi Xuan
2 shared publications
Latest Publications
- Open Access
- 23 May 2026
- Published by Elsevier in SSRN Electronic Journal
Mapping Text to Multiplex Graph: Prompt Compression as Lévy Walk-guided Graph Pruning
Existing prompt compression methods treat text as flat token sequences, failing to capture the distributed nature of important information, which is often spread across multiple locations and connected through both local syntactic dependencies and global semantic relations. Such relational structure is naturally represented as a graph, where tokens or sentences become nodes and their dependencies become edges. To this end, we propose RAGP, which formulates prompt compression as Redundancy-Aware Graph Pruning on a multiplex graph that jointly models fine-grained attention-based dependencies and coarse-grained semantic relations. To efficiently identify non-redundant nodes in this heterogeneous structure (dense local subgraphs and sparse global connections), we employ Lévy walks whose heavy-tailed step distribution naturally balances local exploitation with global exploration. Experiments on LongBench show that RAGP achieves an average score of 49.3 under a 4× compression ratio, outperforming existing LLM-based compression methods, such as LongLLMLingua, which attains 48.8 at a 3× compression ratio. Besides, RAGP also surpasses state-of-the-art vision-based text compression paradigms on multiple tasks.
- 23 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessMapping Text to Multiplex Graph: Prompt Compression as Lévy Walk-guided Graph Pruning
Existing prompt compression methods treat text as flat token sequences, failing to capture the distributed nature of important information, which is often spread across multiple locations and connected through both local syntactic dependencies and global semantic relations. Such relational structure is naturally represented as a graph, where tokens or sentences become nodes and their dependencies become edges. To this end, we propose RAGP, which formulates prompt compression as Redundancy-Aware Graph Pruning on a multiplex graph that jointly models fine-grained attention-based dependencies and coarse-grained semantic relations. To efficiently identify non-redundant nodes in this heterogeneous structure (dense local subgraphs and sparse global connections), we employ Lévy walks whose heavy-tailed step distribution naturally balances local exploitation with global exploration. Experiments on LongBench show that RAGP achieves an average score of 49.3 under a 4× compression ratio, outperforming existing LLM-based compression methods, such as LongLLMLingua, which attains 48.8 at a 3× compression ratio. Besides, RAGP also surpasses state-of-the-art vision-based text compression paradigms on multiple tasks.
- 11 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. PP (00189545) , 1-16
MCLRL: A Multi-Domain Contrastive Learning With Reinforcement Learning Framework for Few-Shot Modulation Recognition
With the rapid development of wireless communication technology, Automatic Modulation Recognition (AMR) is essential for ensuring the security and reliability of communication systems. However, challenges have emerged, including increasing demands for higher device performance and difficulties in data acquisition due to special scenarios. Few-Shot Learning (FSL) provides an effective solution to these challenges. This paper presents a FSL framework for modulation recognition, called MCLRL. In the unsupervised pretraining phase, a combination of multi-representation domain contrastive learning and reinforcement learning enables effective feature extraction. The multi-representation domains of the signals enrich the features, while the reinforcement learning architecture extracts deep features for classification. In the supervised fine-tuning phase, a lightweight attention module and linear classifier are used to filter and classify the features, effectively avoiding overfitting. This approach requires only a few samples and minimal training epochs to achieve strong model performance. The experiments demonstrate that the MCLRL framework effectively extracts key features from the signals, performs well in FSL tasks, and remains flexible in signal model selection.
- 11 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. PP (00189545) , 1-16
MCLRL: A Multi-Domain Contrastive Learning With Reinforcement Learning Framework for Few-Shot Modulation Recognition
With the rapid development of wireless communication technology, Automatic Modulation Recognition (AMR) is essential for ensuring the security and reliability of communication systems. However, challenges have emerged, including increasing demands for higher device performance and difficulties in data acquisition due to special scenarios. Few-Shot Learning (FSL) provides an effective solution to these challenges. This paper presents a FSL framework for modulation recognition, called MCLRL. In the unsupervised pretraining phase, a combination of multi-representation domain contrastive learning and reinforcement learning enables effective feature extraction. The multi-representation domains of the signals enrich the features, while the reinforcement learning architecture extracts deep features for classification. In the supervised fine-tuning phase, a lightweight attention module and linear classifier are used to filter and classify the features, effectively avoiding overfitting. This approach requires only a few samples and minimal training epochs to achieve strong model performance. The experiments demonstrate that the MCLRL framework effectively extracts key features from the signals, performs well in FSL tasks, and remains flexible in signal model selection.
- 24 December 2025
- Published by Elsevier in Information Fusion
- Vol. 130
Vision-language models for person re-identification: a survey and outlook
- 24 December 2025
- Published by Elsevier in Information Fusion
- Vol. 130
Vision-language models for person re-identification: a survey and outlook
- 4 November 2025
- Published by Elsevier in Neurocomputing
- Vol. 662
Dataset distillation with pre-trained models: A contrastive approach
- 4 November 2025
- Published by Elsevier in Neurocomputing
- Vol. 662
Dataset distillation with pre-trained models: A contrastive approach
- Open Access
\emph{FoQuS}: A Forgetting-Quality Coreset Selection Framework for Automatic Modulation Recognition
Deep learning-based Automatic Modulation Recognition (AMR) model has made significant progress with the support of large-scale labeled data. However, when developing new models or performing hyperparameter tuning, the time and energy consumption associated with repeated training using massive amounts of data are often unbearable. To address the above challenges, we propose \emph{FoQuS}, which approximates the effect of full training by selecting a coreset from the original dataset, thereby significantly reducing training overhead. Specifically, \emph{FoQuS} records the prediction trajectory of each sample during full-dataset training and constructs three importance metrics based on training dynamics. Experiments show that \emph{FoQuS} can maintain high recognition accuracy and good cross-architecture generalization on multiple AMR datasets using only 1\%-30\% of the original data.
Open Access\emph{FoQuS}: A Forgetting-Quality Coreset Selection Framework for Automatic Modulation Recognition
Deep learning-based Automatic Modulation Recognition (AMR) model has made significant progress with the support of large-scale labeled data. However, when developing new models or performing hyperparameter tuning, the time and energy consumption associated with repeated training using massive amounts of data are often unbearable. To address the above challenges, we propose \emph{FoQuS}, which approximates the effect of full training by selecting a coreset from the original dataset, thereby significantly reducing training overhead. Specifically, \emph{FoQuS} records the prediction trajectory of each sample during full-dataset training and constructs three importance metrics based on training dynamics. Experiments show that \emph{FoQuS} can maintain high recognition accuracy and good cross-architecture generalization on multiple AMR datasets using only 1\%-30\% of the original data.
- Open Access
DUSE: A Data Expansion Framework for Low-resource Automatic Modulation Recognition based on Active Learning
Although deep neural networks have made remarkable achievements in the field of automatic modulation recognition (AMR), these models often require a large amount of labeled data for training. However, in many practical scenarios, the available target domain data is scarce and difficult to meet the needs of model training. The most direct way is to collect data manually and perform expert annotation, but the high time and labor costs are unbearable. Another common method is data augmentation. Although it can enrich training samples to a certain extent, it does not introduce new data and therefore cannot fundamentally solve the problem of data scarcity. To address these challenges, we introduce a data expansion framework called Dynamic Uncertainty-driven Sample Expansion (DUSE). Specifically, DUSE uses an uncertainty scoring function to filter out useful samples from relevant AMR datasets and employs an active learning strategy to continuously refine the scorer. Extensive experiments demonstrate that DUSE consistently outperforms 8 coreset selection baselines in both class-balance and class-imbalance settings. Besides, DUSE exhibits strong cross-architecture generalization for unseen models.
Open AccessDUSE: A Data Expansion Framework for Low-resource Automatic Modulation Recognition based on Active Learning
Although deep neural networks have made remarkable achievements in the field of automatic modulation recognition (AMR), these models often require a large amount of labeled data for training. However, in many practical scenarios, the available target domain data is scarce and difficult to meet the needs of model training. The most direct way is to collect data manually and perform expert annotation, but the high time and labor costs are unbearable. Another common method is data augmentation. Although it can enrich training samples to a certain extent, it does not introduce new data and therefore cannot fundamentally solve the problem of data scarcity. To address these challenges, we introduce a data expansion framework called Dynamic Uncertainty-driven Sample Expansion (DUSE). Specifically, DUSE uses an uncertainty scoring function to filter out useful samples from relevant AMR datasets and employs an active learning strategy to continuously refine the scorer. Extensive experiments demonstrate that DUSE consistently outperforms 8 coreset selection baselines in both class-balance and class-imbalance settings. Besides, DUSE exhibits strong cross-architecture generalization for unseen models.
- Open Access
ReStNet: A Reusable & Stitchable Network for Dynamic Adaptation on IoT Devices
With the rapid development of deep learning, a growing number of pre-trained models have been publicly available. However, deploying these fixed models in real-world IoT applications is challenging because different devices possess heterogeneous computational and memory resources, making it impossible to deploy a single model across all platforms. Although traditional compression methods, such as pruning, quantization, and knowledge distillation, can improve efficiency, they become inflexible once applied and cannot adapt to changing resource constraints. To address these issues, we propose ReStNet, a Reusable and Stitchable Network that dynamically constructs a hybrid network by stitching two pre-trained models together. Implementing ReStNet requires addressing several key challenges, including how to select the optimal stitching points, determine the stitching order of the two pre-trained models, and choose an effective fine-tuning strategy. To systematically address these challenges and adapt to varying resource constraints, ReStNet determines the stitching point by calculating layer-wise similarity via Centered Kernel Alignment (CKA). It then constructs the hybrid model by retaining early layers from a larger-capacity model and appending deeper layers from a smaller one. To facilitate efficient deployment, only the stitching layer is fine-tuned. This design enables rapid adaptation to changing budgets while fully leveraging available resources. Moreover, ReStNet supports both homogeneous (CNN-CNN, Transformer-Transformer) and heterogeneous (CNN-Transformer) stitching, allowing to combine different model families flexibly. Extensive experiments on multiple benchmarks demonstrate that ReStNet achieve flexible accuracy-efficiency trade-offs at runtime while significantly reducing training cost.
Open AccessReStNet: A Reusable & Stitchable Network for Dynamic Adaptation on IoT Devices
With the rapid development of deep learning, a growing number of pre-trained models have been publicly available. However, deploying these fixed models in real-world IoT applications is challenging because different devices possess heterogeneous computational and memory resources, making it impossible to deploy a single model across all platforms. Although traditional compression methods, such as pruning, quantization, and knowledge distillation, can improve efficiency, they become inflexible once applied and cannot adapt to changing resource constraints. To address these issues, we propose ReStNet, a Reusable and Stitchable Network that dynamically constructs a hybrid network by stitching two pre-trained models together. Implementing ReStNet requires addressing several key challenges, including how to select the optimal stitching points, determine the stitching order of the two pre-trained models, and choose an effective fine-tuning strategy. To systematically address these challenges and adapt to varying resource constraints, ReStNet determines the stitching point by calculating layer-wise similarity via Centered Kernel Alignment (CKA). It then constructs the hybrid model by retaining early layers from a larger-capacity model and appending deeper layers from a smaller one. To facilitate efficient deployment, only the stitching layer is fine-tuned. This design enables rapid adaptation to changing budgets while fully leveraging available resources. Moreover, ReStNet supports both homogeneous (CNN-CNN, Transformer-Transformer) and heterogeneous (CNN-Transformer) stitching, allowing to combine different model families flexibly. Extensive experiments on multiple benchmarks demonstrate that ReStNet achieve flexible accuracy-efficiency trade-offs at runtime while significantly reducing training cost.
- 28 May 2025
- Published by Elsevier in Information Sciences
- Vol. 718
Knowledge-enhanced Relation Graph and Task Sampling for few-shot molecular property prediction
- 28 May 2025
- Published by Elsevier in Information Sciences
- Vol. 718
Knowledge-enhanced Relation Graph and Task Sampling for few-shot molecular property prediction
- Open Access
FCOS: A Two-Stage Recoverable Model Pruning Framework for Automatic Modulation Recognition
With the rapid development of wireless communications and the growing complexity of digital modulation schemes, traditional manual modulation recognition methods struggle to extract reliable signal features and meet real-time requirements in modern scenarios. Recently, deep learning based Automatic Modulation Recognition (AMR) approaches have greatly improved classification accuracy. However, their large model sizes and high computational demands hinder deployment on resource-constrained devices. Model pruning provides a general approach to reduce model complexity, but existing weight, channel, and layer pruning techniques each present a trade-off between compression rate, hardware acceleration, and accuracy preservation. To this end, in this paper, we introduce FCOS, a novel Fine-to-COarse two-Stage pruning framework that combines channel-level pruning with layer-level collapse diagnosis to achieve extreme compression, high performance and efficient inference. In the first stage of FCOS, hierarchical clustering and parameter fusion are applied to channel weights to achieve channel-level pruning. Then a Layer Collapse Diagnosis (LaCD) module uses linear probing to identify layer collapse and removes the collapsed layers due to high channel compression ratio. Experiments on multiple AMR benchmarks demonstrate that FCOS outperforms existing channel and layer pruning methods. Specifically, FCOS achieves 95.51% FLOPs reduction and 95.31% parameter reduction while still maintaining performance close to the original ResNet56, with only a 0.46% drop in accuracy on Sig2019-12. Code is available at https://github.com/yaolu-zjut/FCOS.
Open AccessFCOS: A Two-Stage Recoverable Model Pruning Framework for Automatic Modulation Recognition
With the rapid development of wireless communications and the growing complexity of digital modulation schemes, traditional manual modulation recognition methods struggle to extract reliable signal features and meet real-time requirements in modern scenarios. Recently, deep learning based Automatic Modulation Recognition (AMR) approaches have greatly improved classification accuracy. However, their large model sizes and high computational demands hinder deployment on resource-constrained devices. Model pruning provides a general approach to reduce model complexity, but existing weight, channel, and layer pruning techniques each present a trade-off between compression rate, hardware acceleration, and accuracy preservation. To this end, in this paper, we introduce FCOS, a novel Fine-to-COarse two-Stage pruning framework that combines channel-level pruning with layer-level collapse diagnosis to achieve extreme compression, high performance and efficient inference. In the first stage of FCOS, hierarchical clustering and parameter fusion are applied to channel weights to achieve channel-level pruning. Then a Layer Collapse Diagnosis (LaCD) module uses linear probing to identify layer collapse and removes the collapsed layers due to high channel compression ratio. Experiments on multiple AMR benchmarks demonstrate that FCOS outperforms existing channel and layer pruning methods. Specifically, FCOS achieves 95.51% FLOPs reduction and 95.31% parameter reduction while still maintaining performance close to the original ResNet56, with only a 0.46% drop in accuracy on Sig2019-12. Code is available at https://github.com/yaolu-zjut/FCOS.