Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks162 from public data
- 4,240
- 2,085
- 875
- Foundations of Group Signatures: The Case of Dynamic Groups579
This paper treats the case where the group is static, providing foundations for dynamic group signatures, in the form of a model, strong formal definitions of security, and a construction proven secure under general assumptions.
- Learning Human-to-Humanoid Real-Time Whole-Body Teleoperation311
This work presents Human to Humanoid (H2O), a reinforcement learning (RL) based framework that enables real-time whole-body teleoperation of a full-sized humanoid robot with only an RGB camera, and achieves teleoperation of dynamic whole-body motions in real-world scenarios.
- A Cost-Sensitive Deep Belief Network for Imbalanced Classification284
An evolutionary cost-sensitive deep belief network (ECS-DBN) for imbalanced classification that uses adaptive differential evolution to optimize the misclassification costs based on the training data that presents an effective approach to incorporating the evaluation measure into the objective function.
- A Partition Filter Network for Joint Entity and Relation Extraction172
A partition filter network to model two-way interaction between tasks properly, where feature encoding is decomposed into two steps: partition and filter, and auxiliary experiments suggest that relation prediction is contributory to named entity prediction in a non-negligible way.
- EEG-based estimation of mental fatigue by using KPCA–HMM and complexity parameters154
The investigation suggests that ApEn and Kc can effectively describe the dynamic complexity of EEG, which is strongly correlated with mental fatigue.
- Learning Implicit Sentiment in Aspect-based Sentiment Analysis with Supervised Contrastive Pre-Training149
Supervised Contrastive Pre-training on large-scale sentiment-annotated corpora retrieved from in-domain language resources leads to better capture of both implicit and explicit sentiment orientation towards aspects in reviews, and comprehensive analysis validates its effectiveness on learning implicit sentiment.
- 145
- 120
- Spherical Convolution Empowered Viewport Prediction in 360 Video Multicast with Limited FoV Feedback113
This article proposes a spherical convolution-empowered FoV prediction method, which is a multi-source prediction framework combining salient features extracted from 360-degree video with limited FoV feedback information, and shows that the performance of the proposed method is better than other prediction methods.
- MossFormer2: Combining Transformer and RNN-Free Recurrent Network for Enhanced Time-Domain Monaural Speech Separation103
The integrated MossFormer2 hybrid model demonstrates remarkable enhancements over MossFormer and surpasses other state-of-the-art methods in WSJ0-2/3mix, Libri2Mix, and WHAM!/WHAMR! benchmarks.
- 57
- Automatic recognition of cognitive fatigue from physiological indices by using wavelet packet transform and kernel learning algorithms53
The experimental results show that the predominant activity of autonomic nervous system of subjects turns to the sympathetic activity from parasympathetic activity after the task, and the KPCA-SVM method could be a promising tool for the evaluation of cognitive fatigue.
- 52
- Domain Adaptation Gaze Estimation by Embedding with Prediction Consistency51
This paper proposes an unsupervised method for domain adaptation gaze estimation to eliminate the impact of inter-personal diversity and designs Domain Adaptation Gaze Estimation Network (DAGEN), which learns embedding with prediction consistency and achieves state-of-the-art results on both the MPIIGaze and the EYEDIAP datasets.
- Goal-Guided Generative Prompt Injection Attack on Large Language Models44
This work designed a simple and effective goal-guided generative prompt injection strategy (G2PIA) to find an injection text that satisfies specific constraints to achieve the optimal attack effect approximately and is a query-free black-box attack method with a low computational cost.
- Data security crisis in universities: identification of key factors affecting data breach incidents44
The estimation and robustness findings revealed the underlying mechanisms that impacted university data security, clarifying the understanding of data breaches and suggesting practical implications for universities and other institutes to improve information security.
- FengHuoLun: A Federated Learning based Edge Computing Platform for Cyber-Physical Systems43
A novel Federated Learning based Edge Computing platform for CPS, named “FengHuoLun” is proposed, which can implement smart services where machine learning models are trained in a trusted Federated learning framework and assure the trustworthiness ofSmart services where CPS behaviours are tested and monitored using the Federatedlearning framework.
- 43
- AttackEval: How to Evaluate the Effectiveness of Jailbreak Attacking on Large Language Models37
This work introduces an innovative framework that can help evaluate the effectiveness of jailbreak attacks on LLMs and presents two distinct evaluation frameworks: a coarse-grained evaluation and a fine-grained evaluation.
- DROID: Minimizing the Reality Gap Using Single-Shot Human Demonstration36
The results show that the proposed framework can identify the difference in dynamics between the simulated and the real worlds, and thus improve policy transfer by optimizing the simulator's randomization ranges, and illustrate that based on these same identified parameters, the method can generalize the learned policy to different but related tasks.
- Reinforced Angle-Based Multicategory Support Vector Machines33
It is proved that RAMSVMs can enjoy Fisher consistency, and it is shown that the RAMSVM can be implemented using the very efficient coordinate descent algorithm on its dual problem, using an angle-based prediction rule with k − 1 functions directly.
- 32
- Relative clauses as a benchmark for Minimalist parsing31
It is shown that among those 1600 candidates, a few metrics (and only a few) can provide a unified account of all these contrasts, which significantly limits the number of viable metrics that may be applied to other phenomena, thus reducing theoretical indeterminacy.
- 27
- Attention-Oriented Action Recognition for Real- Time Human-Robot Interaction24
This paper deeply explores the characteristics of the action recognition task in interaction scenes and proposes an attention-oriented multi-level network framework to meet the need for real-time interaction.
- 23
- De’hubert: Disentangling Noise in a Self-Supervised Model for Robust Speech Recognition22
A novel training framework, called deHuBERT, is proposed for noise reduction encoding inspired by H. Barlow’s redundancy-reduction principle, which improves the HuBERT training algorithm by introducing auxiliary losses that drive the self- and cross-correlation matrix between pairwise noise-distorted embeddings towards identity matrix.
- 22
- 20
- 19
- 19
- Invisible Prompts, Visible Threats: Malicious Font Injection in External Resources for Large Language Models18
It is revealed that indirect prompts with injected malicious font can bypass LLM safety mechanisms through external resources, achieving varying success rates based on data sensitivity and prompt design, highlighting the urgent need for enhanced security measures in LLM deployments when processing external content.
- An Edge based Federated Learning Framework for Person Re-identification in UAV Delivery Service17
An edge based FL framework named Fed-UAV is proposed to solve the person reidentification problem in the UAV delivery service which is a typical AI application in smart logistics and can achieve both high accuracy and efficiency in person re-identification while protecting data privacy.
- Contrastive Speech Mixup for Low-Resource Keyword Spotting15
A contrastive speech mixup (CosMix) learning algorithm for low-resource KWS, which introduces an auxiliary contrastive loss to the existing mixup augmentation technique to maximize the relative similarity between the original pre-mixed samples and the augmented samples.
- 15
- Design and FPGA implementation of digital pulse compression for chirp radar based on CORDIC15
The paper presents a full digitized approach for the pulse compression implementation in chirp radars to cancel the quadratic phase term of the echo using a coordinate rotation digital computer (CORDIC).
- Variable Neighborhood Search with Permutation Distance for QAP15
A new concept distance called permutation distance is proposed and exploited in detail and it is shown that this approach would dramatically improve the performance of VNS for QAP.
- Double Perturbation: On the Robustness of Robustness and Counterfactual Bias Evaluation14
A “double perturbation” framework is proposed to uncover model weaknesses beyond the test dataset and is able to reveal the hidden model biases not directly shown in the test datasets.
- Learning Ball-Balancing Robot through Deep Reinforcement Learning14
Simulation results demonstrate that using the compound controller, the ballbot can keep balance under a larger set of initial tilting angles, compared to the conventional model-based controller.
- A Refined Notion of Memory Usage for Minimalist Parsing14
It is shown that none of the difficulty metrics proposed so far can explain why subject relative clauses are more easily processed than object relative clauses in Chinese, Korean, and Japanese, and that a minor tweak to how memory load is determined is sufficient to fully capture the data.
- Estimating VDT Mental Fatigue Using Multichannel Linear Descriptors and KPCA-HMM14
The investigation suggests that the joint KPCA-HMM method can effectively reduce the dimensions of the feature vectors, accelerate the classification speed, and improve the accuracy of mental fatigue to achieve the maximum 88%.
- Learning Highly Dynamic Behaviors for Quadrupedal Robots13
A learning-based approach to enable robots to learn highly dynamic behaviors from animal motion data is proposed and the results show that the controller is able to reproduce highly dynamic behaviors including sprinting, jumping and sharp turning.
- An adaptable active contour model for medical image segmentation based on region and edge information13
Experimental results show that compared with the traditional active contour models, the proposed method can achieve an satisfactory both in segmentation speed and accuracy.
- On the Effect and Remedies of Shrinkage on Classification Probability Estimation13
It is demonstrated that shrinkage creates biases on classification probability estimation, and the proposed refit method gives similar classification accuracy and remarkable improvements on probability estimation on several simulated and real data examples.
- Loss Masking Is Not Needed In Decoder-Only Transformer For Discrete-Token-Based ASR11
This paper proposes a novel approach denoted Smoothed Label Distillation (SLD), which applies a KL divergence loss with smoothed labels on speech tokens, which effectively models speech tokens and outperforms Loss Masking for decoder-only Transformers in ASR tasks with different speech discretization methods.
- Specular highlight removal for endoscopic images using partial attention network11
A specular highlight removal method using a partial attention network (PatNet) for endoscopic imaging is proposed to reduce the interference of bright light in endoscopic surgery and improve the quality of endoscopic Imaging for accurate image analysis.
- Robot path planning based on multi-objective optimization with local search11
A new method with local search is introduced to improve the SPEA2 in this work, which sets up an external population dedicated to local search which can increase the local search ability of the method while retaining good global searching ability.
- An Advanced Tool Wear Forecasting Technique with Uncertainty Quantification Using Bayesian Inference and Support Vector Regression10
A novel method for predicting cutting tool wear using the Brownian motion stochastic process and a SVR model, which shows superiority in prediction accuracy and stability.
- I2CR: Improving Noise Robustness on Keyword Spotting using Inter-Intra Contrastive Regularization10
This work proposed a contrastive regularization method called Inter-Intra Contrastive Regularization (I2CR) to improve the feature representations by guiding the model to learn the fundamental speech information specific to the cluster by maximizing the similarity across Intra and Inter samples of the same class.
- 10
- 10
- HiFi-SR: A Unified Generative Transformer-Convolutional Adversarial Network for High-Fidelity Speech Super-Resolution9
HiFi-SR is a unified network that leverages end-to-end adversarial training to achieve high-fidelity speech super-resolution and significantly outperforms existing speech SR methods across both objective metrics and ABX preference tests, for both in-domain and out-of-domain scenarios.
- 9
- A Novel Estimation Method of Fatigue Using EEG Based on KPCA-SVM and Complexity Parameters9
Compared with several typical classification models, the joint method KPCA-SVM can achieve higher classification accuracy of mental fatigue with less training and classifying time, which indicates that KPCSVM is suitable for the estimation ofmental fatigue.
- 8
- 8
- Fed4ReID: Federated Learning with Data Augmentation for Person Re-identification Service in Edge Computing8
This paper proposes a novel strategy named federated learning with data augmentation for person re-identification (Fed4ReID), which utilises a pre-trained DCGAN (Deep Convolutional Generative Adversarial Network) model forData augmentation at each edge servers to alleviate the impact of Non-IID data.
- Open Domain Chinese Triples Hierarchical Extraction Method8
This paper proposes an integrated open domain Chinese triples hierarchical extraction method (CTHE), considering the advantages of Bi-LSTM-CRF and Att-Bi-GRU models based on the pre-trained BERT encoding model, which can recognize the named entities from Chinese sentences to establish entity pairs, and implement hierarchical extraction of specific and open relations.
- 8
- SPGM: Prioritizing Local Features for Enhanced Speech Separation Performance7
The Single-Path Global Modulation (SPGM) block allows all transformer layers in the model to be dedicated to local feature modelling, making the overall model single-path, and matches the performance of recent SOTA models with up to 8 times fewer parameters.
- Multi-task Prompt Words Learning for Social Media Content Generation7
A new prompt word generation framework based on multi-modal information fusion, which combines multiple tasks including topic classification, sentiment analysis, scene recognition and keyword extraction to generate more comprehensive prompt words is proposed.
- Dual-Memory Multi-Modal Learning for Continual Spoken Keyword Spotting with Confidence Selection and Diversity Enhancement7
This paper proposes a novel framework for replay-based CL in KWS that uses a Dual-Memory Multi-Modal ( DM3 ) structure to enhance generalizability and robustness and introduces a class-balanced selection strategy that uses confidence scores to sort training samples.
- 7
- Exploring Parameter Space with Structured Noise for Meta-Reinforcement Learning7
Experimental results demonstrate the superiority of ESNPS against a number of competitive baselines and utilizes meta-learning and directly uses meta-policy parameters, which contain prior knowledge, as structured noises to perturb the base model for effective exploration in new tasks.
- 7
- FragMangro: A cross-domain zero-shot model for monitoring fragmented mangrove ecosystems6
FragMangro, a cross-domain zero-shot detection model designed to enhance the accuracy and generalization of mangrove detection, serves as an effective tool for ecological monitoring and mangrove conservation, with broader implications for environmental protection and sustainable ecosystem management.
- Online Audio-Visual Autoregressive Speaker Extraction6
This paper proposes a lightweight visual frontend based on depth-wise separable convolution, and proposes a lightweight autoregressive acoustic encoder to serve as the second cue, to actively explore the information in the separated speech signal from past steps.
- Online Social Behavior Enhanced Detection of Political Stances in Tweets6
This paper constructs a bipartite graph that explicitly models user-tweet interactions, which provides a comprehensive contextual understanding of the 2020 US presidential election and introduces a novel skip aggregation mechanism to effectively leverage the wealth of user behavioral information encoded in this graph.
- Adaptive Knowledge Distillation Between Text and Speech Pre-Trained Models6
This paper proposes the Prior-informed Adaptive knowledge Distillation (PAD) that adaptively leverages text/speech units of variable granularity and prior distributions to achieve better global and local alignments between text and speech pre-trained models.
- A Web News Classification Method: Fusion Noise Filtering and Convolutional Neural Network6
A web news classification method via fusing noise detection, BERT-based semantic similarity noise filtering and convolutional neural network (NF-CNN) to solve the problem of noisy web news content classification.
- 6
- Adapter-tuning with Effective Token-dependent Representation Shift for Automatic Speech Recognition5
This approach adds a token-specific representation shift (bias) to the intermediate representations of a pre-trained model, which better maps the latent features of the frozen network to the task domain.
- 5
- Fusion of Domain Knowledge and Text Features for Query Expansion in Citation Recommendation5
Experiments of citation recommendation for papers in public data sets show that the proposed model of query expansion improves the performance of academic citation recommendation.
- The candidate multi-cut for cell segmentation5
This paper introduces a model, the Candidate Multi-Cut (CMC), that allows joint selection and clustering of segment candidates from a merge-tree and solves the optimization problem of selecting and clustered of candidates using an integer linear program.
- Temporal Query Processing in Social Network5
This paper mainly study three kinds of temporal queries in social network, which explore temporal attribute in user’s online duration, friendship duration and participation in social activities, and designs query algorithms to solve the three queries.
- 4
- 4
- 4
- Estimating Mental Fatigue Based on Multichannel Linear Descriptors Analysis4
The experimental results show that the average values of Ω, Φ and Σ of EEG decreases with the prolonging of sleep deprivation and working time, and then also decrease with the increase of mental fatigue level.
- Are Soft Prompts Good Zero-Shot Learners for Speech Recognition?3
The role of soft prompts in automatic speech recognition (ASR) is investigated and an effective modification on noise prompts is proposed to show that they are capable of zero-shot learning on adapting to out-of-distribution noise environments.
- 3
- Dual Acoustic Linguistic Self-supervised Representation Learning for Cross-Domain Speech Recognition3
This work proposes an end-to-end acoustic and linguistic integrated representation learning model, namely Dual-w2v-BART, that incorporates acoustic representations from wav2vec2.0 and linguistic information from BART model by utilizing the cross-attention mechanism in the decoder, with paired speech-text dual inputs.
- 3
- 3
- RF2Net: salient object detection using level set loss and reverse attention fusion feed network3
The proposed RF2Net incorporates the idea of level set structured loss and reverse attention mechanism on the basis of F3Net and shows that the proposed algorithm can effectively improve the effect of salient object detection.
- 3
- 3
- Webpage visual feature extraction and similarity algorithm3
A visual feature extraction method and a visual similarity algorithm for better identifying phishing websites that integrates the similarity of the visual blocks similarity and visual block weight to get the web pages' visual similarity.
- Color image enhancement algorithm based on edge extraction3
To solve the problem of blurred edges in some color images, a color image enhancement algorithm based on edge extraction based on improved Laplace operator is proposed, which improves the clarity of edge while preserving image color.
- 3
- Image segmentation model based on adaptive adjustment of global and local information3
An adaptive mixture model for image segmentation that synthesizes both global information and local information using a new adaptive balance function has been proposed and can adaptively adjust the weighting to drive curve evolution trends and states.
- 2
- A Simple and Effective Baseline for Attentional Generative Adversarial Networks2
Improvements are made to remove redundancy structure and improve the backbone network of AttnGAN to integrate and reconstruct multiple losses of DAMSM and proposed the SEAttnGAN, which significantly improved the model size and training efficiency.
- A Feature Extracting and Matching System Based on Magic-Number and AC-Algorithm2
Through the analysis of the file features proposed in this paper, a lot of repetitive work can be reduced and the efficiency of traffic analysis can be improved.
- Tourism Forecast Based on Web Search Data and Sentiment Analysis of Social Network2
This essay studies the tourist volume in Sanya, China from January 2012 to December 2019, depending on web search data and social media data, and builds ARIMA model based on search index and sentiment index, to predict monthly touristVolume in the near future.
- MEMORY BASED DECISION MAKING: A SPIKING NEURAL CIRCUIT MODEL2
A biologically-based network model of spiking neurons for competition between automatic response and conscious decision making and uses episodic memory output as the top-down input of decision making can effectively explain the transmission mechanism of conscious decision information.
- Rule-Based HierarchicalRank: An Unsupervised Approach to Visible Tag Extraction from Semi-structured Chinese Text2
This paper proposes a method to automatically extract a small set of visible tag from a document that can accurately reveal the document’s topic and can facilitate fast information processing.
- 2
- 2
- REC: fast sparse regression-based multicategory classification2
A new sparse REgression based multicategory Clas-sifier (REC), which uses a simplex to represent differ-ent categories of the classification problem and enjoys an extraordinarily fast computational speed.
- [Mental fatigue electroencephalogram signals analysis based on singular system].2
The experimental results showed that the contribution of the largest principal component of EEG signals increased in the prefrontal, frontal and central areas, while the number of principal component needed for accumulative contribution decreased by 95% with the increasing mental fatigue level.
- 1
- A Novel ViDAR Device With Visual Inertial Encoder Odometry and Reinforcement Learning-Based Active SLAM Method1
A novel visual-inertial-encoder tightly coupled odometry (VIEO) based on a video detection and ranging (ViDAR) device and a platform motion decoupled active SLAM method based on deep reinforcement learning (DRL) is proposed.
- Comparative Genomics Analysis Provides Insights into Pregnancy Maintenance Mechanisms in the Ovoviviparous Teleost Black Rockfish (Sebastes schlegelii)1
A comparative genomic analysis between S. schlegelii and representative oviparous and ovoviviparous fishes identifies 374 lineage-specific gene families associated with cellular signaling, metabolic regulation, immune response, and developmental processes that contribute to a broader understanding of the genetic adaptations with pregnancy in ovoviviparous teleosts.
- 1
- 1
- 1
- 1
- 1
- 1
- Instant Messaging Application Traffic Recognition1
A traffic recognition method based on deep learning and machine learning is implemented for unsupervised clustering of traffic on the UNB ISCX data set and the experimental results are quite good.
- 1
- 1
- 1
- 1
- Historical Geo-Social Query Processing1
This paper addresses two historical geo-social queries on group problem, one for check-in historical query and the other for life pattern query, and builds sophisticated structures to index the data and proposes algorithms to process queries.
- Improbable Differential Attacks on Reduced FOX641
5-round improbable differentials of FOX are found and used to attack 6, 7, and 8-round FOX64 and it is proved that it is possible to obtain differentials such that the predicted differences occur less frequently for the correct key.
- 1
- –
- –
- –
- –
- –
- Data job requirement characteristics analysis based on large language model and K-means clustering–
The research found that the quality of key attribute extraction using ChatGLM3 improved by 54.1% compared to dictionary-based methods, significantly reducing the workload of manual annotation.
- –
- –
- –
- A Unified Recognition and Correction Model under Noisy and Accent Speech Conditions–
Experimental results on accented and noisy speech datasets demonstrate that the proposed UAC framework effectively produces improved word error rate when compared against the pipeline baselines.
- Bridging the Projection Gap: Overcoming Projection Bias Through Parameterized Distance Learning–
This work improved the network structure of VAEGAN to leverage the discriminative models of two branches to separately predict the seen samples and the unseen samples generated by this seen one and proposed a new loss function with two branches to help us learn the optimized Mahalanobis distance representation.
- –
- –
- Design and Implementation of Information Detection System Based on Privacy Sensitive Data–
This paper mainly studies the detection of private data by collecting data traffic in the network, analyzing the content, and determining whether the data is encrypted during the transmission process.
- –
- –
- –
- –
- –
- Rejoinder of “Double sparsity kernel learning with automatic variable selection and data extraction”–
- Method for improving zero‐shot image classification–
Experiments show that zero-shot image classification based on Pearson correlation coefficient is better than Euclidean distance and cosine similarity.
- –
- –
- –
- –
- –
- –
- –
- A Re-sampling Low-pass Filter for Differential Protection Designed with Artificial Neural Network–
This method can help to design FIR filters with any amplitude or phase in frequency domain and can reduce the differential current error by carrying out frequency domain compensation on differential current according to the characteristics of leading low-pass filters and improve the reliability of the protection.
- –
- –
- Develop of Master Station Facility Based on the Distributed Structure of the Oil with TWACS Well Remote Monitoring System–
A TWACS master facility which based on the work of the distributed structure to Meet well remote monitoring and management is designed and implemented.
- –
- –
- Removing of Preprinted Characters in Document Image Processing–
A practical preprinted character removing method that is efficient on real-life check images, and based on the varying of stroke contours.
- An Efficient Geometric Distortions Correction Arithmetic for Color Image Digital Watermarking–
A novel geometric distortions correction schema for image is proposed, it takes the advantage of the corner center angle and corner certer distance firstly to acquire geometric distort correction.
- Medical Image Compression Based on Quadtree Wavelet Fractal–
The test has proved that the new algorithm of medical image compression based on quadtree wavelet fractal can achieve better compression without diagnostic loss and can be potentially used in telemedicine and PACS.
- [Digital watermarking for electroencephalogram compression].–
The current state of digital watermarking and EEG compression inside and outside country is summarized and future endeavors are discussed.
- ALinear Transformation Model on Face Image and ItsApplication in Recognition–
A linear transformation model on face image is proposed that is feasible, simple and tractable in computing, and very effective to weaken the aftereffect of varying illumination.
- The reform of PE in Nagoya university of Japan–
The character of the reform is analyzed, which are the clear reform idea, the course design of thinking much of theories studying about health and gymnastic sciences, and the active teaching form in exciting students' ability.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.