Author name cluster

Dan Liu

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.

12 papers

2 author rows

AAAI Conference 2026 Conference Paper

Causal-Tune: Mining Causal Factors from Vision Foundation Models for Domain Generalized Semantic Segmentation

Yin Zhang
Yongqiang Zhang
Yaoyue Zheng
Bogdan Raducanu
Dan Liu

Fine-tuning Vision Foundation Models (VFMs) with a small number of parameters has shown remarkable performance in Domain Generalized Semantic Segmentation (DGSS). Most existing works either train lightweight adapters or refine intermediate features to achieve better generalization on unseen domains. However, they both overlook the fact that long-term pre-trained VFMs often exhibit artifacts, which hinder the utilization of valuable representations and ultimately degrade DGSS performance. Inspired by causal mechanisms, we observe that these artifacts are associated with non-causal factors, which usually reside in the low- and high-frequency components of the VFM spectrum. In this paper, we explicitly examine the causal and non-causal factors of features within VFMs for DGSS, and propose a simple yet effective method to identify and disentangle them, enabling more robust domain generalization. Specifically, we propose Causal-Tune, a novel fine-tuning strategy designed to extract causal factors and suppress non-causal ones from the features of VFMs. First, we extract the frequency spectrum of features from each layer using the Discrete Cosine Transform (DCT). A Gaussian band-pass filter is then applied to separate the spectrum into causal and non-causal components. To further refine the causal components, we introduce a set of causal-aware learnable tokens that operate in the frequency domain, while the non-causal components are discarded. Finally, refined features are transformed back into the spatial domain via inverse DCT and passed to the next layer. Extensive experiments conducted on various cross-domain tasks demonstrate the effectiveness of Causal-Tune. In particular, our method achieves superior performance under adverse weather conditions, improving +4.8% mIoU over the baseline in snow conditions.

PDF Details DOI

JBHI Journal 2026 Journal Article

RE-HPBS-IPIC: A Resting EEG- and High-Activation Pain Brain Source-Driven Framework for Inter-Subject Pain Intensity Classification

Wenjia Gao
Dan Liu
Qisong Wang
Yongping Zhao
Jinwei Sun

Objective: Accurate inter-subject pain intensity assessment using EEG remains a major challenge due to substantial inter-subject variability. This study introduces a novel framework that leverages pain-related brain dynamics and transfer learning to enable reliable inter subject pain intensity classification. Methods: The pro-posed method first quantifies pain sensitivity from resting state EEG to identify source subjects with comparable neural pain signatures. High-activation pain brain sources are subsequently localized and remapped between source and target subjects. A classifier is trained to evaluate transfer suitability across subjects, and balanced distribution adaptation is applied to align brain source features, mitigating inter-subject variability. The adapted model infers pseudo labels for the target EEG, which guide the pain response extraction. Final classification is determined by selecting the model exhibiting the minimal cross-domain discrepancy between brain source and pain-evoked EEG features. Results: Experimental evaluations on real EEG datasets demonstrate that the proposed method significantly out performs three existing approaches in inter-subject pain intensity classification. Significance: The proposed method effectively overcomes the problem of poor reliability in inter-subject pain intensity classification, providing a robust and clinically viable solution.

Details DOI

EAAI Journal 2025 Journal Article

Deep reinforcement learning for optimization of spiral shaft design in shield machine

Lin Lin
Lingyu Yue
Song Fu
Dan Liu
Yancheng Lv
Yikun Liu
Sihao Zhang
Shiwei Suo

The increasing demand for highly customized shield machines imposes greater efficiency and accuracy requirements on the optimization. To address the inefficiencies and susceptibility to local optima in existing structural design methods, a novel approach to construct a simulation analysis surrogate model for creating an environment for the optimization of spiral shaft is introduced. It also improves the Deep Deterministic Policy Gradient (DDPG) algorithm to optimize the structural parameters. In the surrogate model, a Goal-oriented autoencoder (GOAE)-classifier model is employed to discriminate the feasibility of samples, and a hybrid surrogate model, utilizing a Self-attention Artificial Neural Network (Self-attention ANN) weight allocation mechanism, makes precise predictions for feasible samples. This model automatically assigns adaptive weights to sub-surrogate models. Within the DDPG framework, a novel serial-parallel hybrid structure for the Actor network is proposed, harnessing the specialized feature representation capabilities of multiple networks to enhance optimization policy accuracy. Additionally, an experience replay filtering mechanism based on sample similarity is introduced to ensure sample diversity and boost the performance of the optimization policy. A simulation analysis surrogate model is constructed on a generated dataset, and the improved DDPG algorithm is leveraged for the optimization of spiral shafts based on this surrogate model. Experimental results demonstrate that the constructed simulation analysis surrogate model facilitates rapid and precise analysis of spiral shaft, while the improved DDPG algorithm successfully optimizes spiral shaft structural parameters, ultimately improving spiral shaft performance.

Details DOI

JBHI Journal 2025 Journal Article

EEG-DG: A Multi-Source Domain Generalization Framework for Motor Imagery EEG Classification

Xiao-Cong Zhong
Qisong Wang
Dan Liu
Zhihuang Chen
Jing-Xiao Liao
Jinwei Sun
Yudong Zhang
Feng-Lei Fan

Motorimagery EEG classification plays a crucial role in non-invasive Brain-Computer Interface (BCI) research. However, the performance of classification is affected by the non-stationarity and individual variations of EEG signals. Simply pooling EEG data with different statistical distributions to train a classification model can severely degrade the generalization performance. To address this issue, the existing methods primarily focus on domain adaptation, which requires access to the test data during training. This is unrealistic and impractical in many EEG application scenarios. In this paper, we propose a novel multi-source domain generalization framework called EEG-DG, which leverages multiple source domains with different statistical distributions to build generalizable models on unseen target EEG data. We optimize both the marginal and conditional distributions to ensure the stability of the joint distribution across source domains and extend it to a multi-source domain generalization framework to achieve domain-invariant feature representation, thereby alleviating calibration efforts. Systematic experiments conducted on a simulative dataset, BCI competition IV 2a, 2b, and OpenBMI datasets, demonstrate the superiority and competitive performance of our proposed framework over other state-of-the-art methods. Specifically, EEG-DG achieves average classification accuracies of 81. 79% and 87. 12% on datasets IV-2a and IV-2b, respectively, and 78. 37% and 76. 94% for inter-session and inter-subject evaluations on dataset OpenBMI, which even outperforms some domain adaptation methods.

Details DOI

EAAI Journal 2025 Journal Article

Recent advances in flotation froth image analysis via deep learning

Xin Chen
Dan Liu
Longzhou Yu
Ping Shao
Mingyan An
Shuming Wen

Flotation froth image analysis with computer vision systems has witnessed a transformative evolution through the integration of deep learning. Deep learning outperforms traditional feature design by effectively learning intricate feature representations, thus enhancing the assessment of froth flotation processes' operational performance. Flotation froth image analysis via deep learning facilitates real-time monitoring of dynamic flotation processes, guiding the adjustment of operational variables through predicting performance indicators, recognizing froth states and segmenting foam edges, which promotes resource efficiency and supports the sustainable development of beneficiation. Despite the vast potential of deep learning for time-series forecasting within the multistage flotation cycle, its capabilities remain underexplored. To fill this gap, based on recent research, we discuss the application of temporal and multistage information in flotation cycle. We introduce the development trends of deep learning in various processes of flotation froth image analysis, including data collection, dataset preprocessing, feature extraction, and modeling. We particularly discuss advanced techniques for extracting time-series features, and developing multistage models and innovative data collection methods, so as to emphasize the importance of using temporal information. Eventually, the review explores several trends and challenges for future research. This review is expected to leave readers with deeper thoughts about algorithm design and data collection in the flotation domain, thereby promoting further research and development in beneficiation automation.

Details DOI

EAAI Journal 2023 Journal Article

An integrating spherical fuzzy AHP and axiomatic design approach and its application in human–machine interface design evaluation

Qinghua Liu
Jiadui Chen
Kai Yang
Dan Liu
Ling He
Qing Qin
Yuqing Wang

Human–machine interface (HMI) design evaluation is critical in interactive product development because it directly affects the cost of subsequent design and user experiments. The evaluation information of HMI design mainly depends on the subjective perception and preference of experts, especially the hesitancy degree is rarely considered. We propose an integrated spherical fuzzy AHP (SF-AHP) and spherical fuzzy axiomatic design (SF-AD) method to choose a reasonable HMI alternative, considering the potential risks caused by the hesitancy degree of experts. Firstly, we introduce a process of automatically repairing inconsistent spherical fuzzy preference relations (SFPR) for SF-AHP. After that, we build suitable evaluation criteria and calculate the criteria weight by the SF-AHP method. We further extend the axiomatic design to the spherical fuzzy environment and propose a multi-criteria decision-making (MCDM) method based on SF-AD to evaluate HMI alternatives. The case analysis results demonstrate the effectiveness of the proposed method in the HMI design evaluation process, and the sensitivity and comparative analysis results indicate that the proposed method is stable and reliable.

Details DOI

JBHI Journal 2023 Journal Article

FBLPF-ABOW: An Effective Method for Blink Artifact Removal in Single-Channel EEG Signal

Wenjia Gao
Dan Liu
Qisong Wang
Yongping Zhao
Jinwei Sun

Objective: The latest development in low-cost single-channel Electroencephalography (EEG) devices is gaining widespread attention because it reduces hardware complexity. Discrete wavelet transform (DWT) has been a popular solution to eliminate the blink artifacts in EEG signals. However, the existing DWT-based methods share the same wavelet function among subjects, which ignores the individual difference. To remedy this deficiency, this article proposes a novel approach to eliminate the blink artifacts in single-channel EEG signals. Methods: Firstly, the forward-backward low-pass filter (FBLPF) and a fixed-length window are used to detect blink artifact intervals. Secondly, the adaptive bi-orthogonal wavelet (ABOW) is constructed based on the most representative blink signal. Thirdly, these detected signals are filtered by ABOW-DWT. The DWT's decomposition depth is automatically chosen by a similarity-based method. Results: Compared to eight state-of-the-art methods, experiments on semi-simulated and real EEG signals demonstrate the proposed method's superiority in removing the blink artifacts with less neural information loss. Significance: To filter the blink artifacts in single-channel EEG signals, the innovative idea of constructing an adaptive wavelet function based on the signal characteristics rather than using the conventional wavelet is proposed for the first time.

Details DOI

EAAI Journal 2023 Journal Article

Multi-agent quantum-inspired deep reinforcement learning for real-time distributed generation control of 100% renewable energy systems

Dan Liu
Yingzi Wu
Yiqun Kang
Linfei Yin
Xiaotong Ji
Xinghui Cao
Chuangzhi Li

With promoting peaking carbon emissions and achieving carbon neutrality, the real-time distributed control of the prosumers of 100% renewable energy systems (RESs) is challenging. This paper proposes multi-agent quantum-inspired deep reinforcement learning (QDRL) approaches for real-time distributed generation control of 100% RESs. Quantum-inspired operation is introduced into deep reinforcement learning (DRL) as quantum-inspired Q-learning, quantum-inspired state–action–reward-state–action, quantum-inspired deep Q-network, quantum-inspired policy gradient, quantum-inspired deep deterministic policy gradient, quantum-inspired twin-delayed deep deterministic policy gradient, quantum-inspired actor–critic, quantum-inspired proximal policy optimization, and quantum-inspired soft actor–critic. These proposed nine QDRL approaches are compared with DRL approaches under two 100% RESs. The numeric results show that the QDRL obtains more minor carbon emissions and frequency deviations under complex 100% RESs. Moreover, the quantum states of QDRL match the uncertain states of the prosumers of 100% RESs. Besides, the exploration and exploitation of the QDRL for the real-time control problems of multi-agent systems are verified and analyzed.

Details DOI

ICRA Conference 2023 Conference Paper

Weakly Supervised Referring Expression Grounding via Target-Guided Knowledge Distillation

Jinpeng Mi
Song Tang 0001
Zhiyuan Ma 0001
Dan Liu
Qingdu Li
Jianwei Zhang 0001

Weakly supervised referring expression grounding aims to train a model without the manual labels between image regions and referring expressions during the training phase. Current predominant models often adopt deep structures to reconstruct the region-expression correspondence. A crucial deficiency of the existing approaches lies in that these models neglect to exploit potential valuable information to further improve their grounding performance. To address this issue, we leverage knowledge distillation as a unique scheme to excavate and transfer helpful information for acquiring a better model. Specifically, we propose a target-guided knowledge distillation framework that accounts for region-expression pairs reconstruction and matching. We reactivate the target-related prediction information learned by a pre-trained teacher model and transfer the target-related prediction knowledge from the teacher to guide the training process and boost the performance of the student model. We conduct extensive experiments on three benchmark datasets, i. e. , RefCOCO, RefCOCO+, and RefCOCOg. Without bells and whistles, our approach achieves state-of-the-art results on several splits of benchmark datasets. The implementation codes and trained models are available at: https://github.com/dami23/WREG_KD.

Details

AAAI Conference 2019 Conference Paper

MPD-AL: An Efficient Membrane Potential Driven Aggregate-Label Learning Algorithm for Spiking Neurons

Malu Zhang
Jibin Wu
Yansong Chua
Xiaoling Luo
Zihan Pan
Dan Liu
Haizhou Li

One of the long-standing questions in biology and machine learning is how neural networks may learn important features from the input activities with a delayed feedback, commonly known as the temporal credit-assignment problem. The aggregate-label learning is proposed to resolve this problem by matching the spike count of a neuron with the magnitude of a feedback signal. However, the existing threshold-driven aggregate-label learning algorithms are computationally intensive, resulting in relatively low learning efficiency hence limiting their usability in practical applications. In order to address these limitations, we propose a novel membrane-potential driven aggregate-label learning algorithm, namely MPD-AL. With this algorithm, the easiest modifiable time instant is identified from membrane potential traces of the neuron, and guild the synaptic adaptation based on the presynaptic neurons’ contribution at this time instant. The experimental results demonstrate that the proposed algorithm enables the neurons to generate the desired number of spikes, and to detect useful clues embedded within unrelated spiking activities and background noise with a better learning efficiency over the state-of-the-art TDP1 and Multi-Spike Tempotron algorithms. Furthermore, we propose a data-driven dynamic decoding scheme for practical classification tasks, of which the aggregate labels are hard to define. This scheme effectively improves the classification accuracy of the aggregate-label learning algorithms as demonstrated on a speech recognition task.

PDF Details

AAAI Conference 2019 Conference Paper

Scale Invariant Fully Convolutional Network: Detecting Hands Efficiently

Dan Liu
Dawei Du
Libo Zhang
Tiejian Luo
Yanjun Wu
Feiyue Huang
Siwei Lyu

Existing hand detection methods usually follow the pipeline of multiple stages with high computation cost, i. e. , feature extraction, region proposal, bounding box regression, and additional layers for rotated region detection. In this paper, we propose a new Scale Invariant Fully Convolutional Network (SIFCN) trained in an end-to-end fashion to detect hands efficiently. Specifically, we merge the feature maps from high to low layers in an iterative way, which handles different scales of hands better with less time overhead comparing to concatenating them simply. Moreover, we develop the Complementary Weighted Fusion (CWF) block to make full use of the distinctive features among multiple layers to achieve scale invariance. To deal with rotated hand detection, we present the rotation map to get rid of complex rotation and derotation layers. Besides, we design the multi-scale loss scheme to accelerate the training process significantly by adding supervision to the intermediate layers of the network. Compared with the state-of-the-art methods, our algorithm shows comparable accuracy and runs a 4. 23 times faster speed on the VIVA dataset and achieves better average precision on Oxford hand detection dataset at a speed of 62. 5 fps.

PDF Details

ICRA Conference 2005 Conference Paper

Haptic Device with Gripping Force Feedback

Qunzhi Li
Shuxin Wang
Jintian Yun
Dan Liu
Baoping Han

This paper presents a haptic device with gripping force feedback, which is a part of the microsurgery robot system developed in our lab. This device includes a master manipulator with an interactive gripping force sensing, and a relative sensing finger located at the end of the slave manipulator. Their effectiveness is tested and verified by experiments. It can be concluded from an animal experiment that this haptic device can realize the interactive gripping force sensing under the master-slave manipulation.

Details