Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks94 from public data
- 245
- Constructing Self-Motivated Pyramid Curriculums for Cross-Domain Semantic Segmentation: A Non-Adversarial Approach239
This work proposes a new approach, called self-motivated pyramid curriculum domain adaptation (PyCDA), to facilitate the adaptation of semantic segmentation neural networks from synthetic source domains to real target domains and reports state-of-the-art results for the adaptation from both GTAV and SYNTHIA to Cityscapes.
- Can Cross Entropy Loss Be Robust to Label Noise?220
A neural network that learns the relationships among the predictors associated to each class, yielding First-Order Logic (FOL)-based descriptions is proposed, which can integrate human-driven preferences that guide the learning-to-explain process, and it is presented in a unified framework.
- An efficient instance selection algorithm to reconstruct training set for support vector machine148
This paper focuses on the fact of uneven distribution of instances in the vector space to propose an efficient self-adaption instance selection algorithm from the viewpoint of geometry-based method and results show that the proposed algorithm outperforms most of competitive algorithms due to its high efficiency and efficacy.
- 102
- 82
- 78
- Spatiotemporal Variations of Air Pollutants and Ozone Prediction Using Machine Learning Algorithms in Beijing-Tianjin-Hebei Region from 2014 to 202169
A seasonal and diurnal pattern with summer and afternoon peaks was found for ozone, which was contrast with other pollutants, and RF model was beneficial and suggested for predicting spatiotemporal variations of ozone in future studies.
- An Effective Conversation‐Based Botnet Detection Method61
The experimental results show the conversation-based detection approach can identify botnet with higher accuracy and lower false positive rate than flow-based approach.
- 49
- C 2 IMUFS: Complementary and Consensus Learning-Based Incomplete Multi-View Unsupervised Feature Selection45
A complementary and consensus learning-based incomplete multi-view unsupervised feature selection method (IMUFS), which integrates feature selection into an extended weighted non-negative matrix factorization model equipped with adaptive learning of view-weights and a sparse sparse linear combinations of multiple similarity matrices derived from different views.
- A new Centroid-Based Classification model for text categorization43
The proposed gravitation model is proved to be less over-fitting and has higher learning ability than CBC model, which shows that it consistently outperforms CBC together with the Class-Feature-Centroid Classifier (CFC).
- Adaptive graph-based generalized regression model for unsupervised feature selection40
A novel generalized regression model equipped with an uncorrelated constraint and the l 2 , 1 -norm regularization is presented, which can simultaneously select the uncor related and discriminative features as well as reduce the variance of these data points belonging to the same neighborhood, which is helpful for the clustering task.
- 39
- Breast tumor classification through learning from noisy labeled ultrasound images37
This work proposes an effective approach called noise filter network (NF-Net) to address the problem of noisy labels when training breast tumor classification models, and can generate benign/malignant labels according to the BI-RADS ratings.
- TRRNet: Tiered Relation Reasoning for Compositional Visual Question Answering36
A novel tiered reasoning method that dynamically selects object level candidates based on language representations and generates robust pairwise relations within the selected candidate objects and achieves state-of-the-art performance on two VQA datasets.
- 33
- Long short‐term memory on abstract syntax tree for SQL injection detection32
A novel detection approach based on long short ‐ term memory and abstract syntax tree is proposed, which could detect SQLIAs from the raw query strings and work under SQL detection bypassing scenario.
- Colorectal Polyp Segmentation in the Deep Learning Era: A Comprehensive Survey30
This paper provides a comprehensive review of the current deep CPS with a novel taxonomy, including network architectures, level of supervision, and learning paradigm, and discussed deep learning-based CPS methods' main challenges and opportunities.
- Learning cross-domain semantic-visual relationships for transductive zero-shot learning30
The Transferrable Semantic-Visual Relation (TSVR) approach towards transductive ZSL is proposed, which redefines image recognition as predicting the similarity/dissimilarity labels for semantic-visual fusions consisting of class attributes and visual features.
- TarGAN: Generating target data with class labels for unsupervised domain adaptation28
A new unsupervised domain adaptation method named TarGAN is proposed, which can successfully disentangle the class and the style codes of the target generator and generate Target samples with given class labels based on the recent advances of Generative Adversarial Networks.
- 26
- Partial Label Learning with Semantic Label Representations26
A novel framework partial label learning with semantic label representations dubbed ParSE is proposed, which consists of two synergistic processes, including visual-semantic representation learning and powerful label disambiguation.
- 25
- Online COVID-19 diagnosis with chest CT images: Lesion-attention deep neural networks24
A lesion-attention deep neural network (LA-DNN) is developed to predict COVID-19 positive or negative with a richly annotated chest CT image dataset, and the experimental results show that the area under the curve (AUC) and sensitivity for the diagnosis of CO VID-19 patients are 91.2% and 85.7% respectively, which reach the clinical standards for practical use.
- Learning Unbiased Zero-Shot Semantic Segmentation Networks Via Transductive Transfer23
This letter proposes an easy-to-implement transductive approach to alleviate the prediction bias in zero-shot semantic segmentation, and demonstrates the effectiveness of this approach over the PASCAL dataset.
- Self-Training Vision Language BERTs With a Unified Conditional Model22
A self-training approach that allows training VL-BERTs from unlabeled image data and is able to get competitive or even better performances compared to the models of similar model size trained with 3 million extra image data.
- 20
- 16
- 15
- 14
- 13
- 12
- Incorporating multiple cluster centers for multi-label learning12
This paper proposes a novel data augmentation approach that performs clustering on the real examples and treats the cluster centers as virtual examples, and these virtual examples naturally embody the local label correlations and label importances.
- Integrating Traffics with Network Device Logs for Anomaly Detection12
Inspired by multiplatform intrusion detection techniques, TLCD correlates log data with traffic characteristics to reflect the attack process and construct a federated detection platform that works well with high accuracy and low false positive rate.
- The convergence and termination criterion of quantum-inspired evolutionary neural networks12
An appropriate termination criterion based on the average convergence rate (ACR) can duly stop the training process of QENN and overcome the limitations of the termination criterionbased on the probability of generating the best solution (PBS).
- 10
- 10
- 9
- 9
- HLV-1K: A Large-scale Hour-Long Video Benchmark for Time-Specific Long Video Understanding7
This paper focuses on building a large-scale hour-long long video benchmark, HLV-1K 1, designed to evaluate long video understanding models, and evaluates its value for testing deep long video understanding capabilities at different levels and for various tasks.
- 7
- 7
- 7
- 7
- Latent Gaussian process for anomaly detection in categorical data7
This work proposes a semi-supervised approach towards anomaly detection in multivariate categorical data that learns the probability distribution of normal instances with the assumption that the categoricalData are generated from a continuous latent space.
- 7
- Generative classification model for categorical data based on latent Gaussian process7
A generative classification model as a supervised method for labelled categorical data is proposed, in which latent gaussian process is used to estimate the class-conditional densities, and can achieve better classification performance compared with other classification models for categoricalData.
- Sentiment-oriented Sarcasm Integration for Video Sentiment Analysis Enhancement with Sarcasm Assistance6
The Progressively Sentiment-oriented Sarcasm Refinement and Integration (PS2RI) framework, which focuses on modeling sentiment-oriented sarcasm features to enhance sentiment prediction and iteratively performs sentiment-oriented sarcasm refinement and sarcasm integration operations within the sentiment recognition framework.
- 6
- Curriculum Knowledge Distillation for Emoji-supervised Cross-lingual Sentiment Analysis6
This work proposes a novel cross-lingual sentiment analysis approach dubbed Curriculum Knowledge Distiller (CKD), which uses emojis to bridge the source and target languages and distills ISCs into the Target Language Sentiment Classifier (TSC) following the curriculum learning mechanism.
- 6
- Large Language Models Meet Contrastive Learning: Zero-Shot Emotion Recognition Across Languages5
This paper proposes leveraging contrastive learning to refine multilingual speech features and extend large language models for zero-shot multilingual speech emotion estimation, and introduces a large-scale synthetic multilingual speech emotion dataset, M5SER.
- Defect detection in railway catenary systems: a survey5
This survey provides a comprehensive overview of recent progress in intelligent defect detection for railway catenary infrastructure, and classify existing approaches into four major model categories and examine the key techniques employed within each.
- Rethinking the Effect of Unimodal Labels in Multimodal Sentiment Analysis4
The proposed Multimodal Task Correlation-aware Learning (MTCL) framework introduces a Correlation-Adaptive Training (CAT) strategy to learn a task-relation aware unimodal encoder for each modality, which can effectively distinguish beneficial and harmful task correlations to extract distinctive and robust unimodal representations.
- 4
- 4
- Attribute-formed Class-specific Concept Space: Endowing Language Bottleneck Model with Better Interpretability and Scalability3
The Attribute-formed Language Bottleneck Model (ALBM) is proposed, which organizes concepts in the attribute-formed class-specific space, where concepts are descriptions of specific attributes for specific classes.
- 3
- Why is a Bird's Caption a Good Demonstration? Towards Effective Multimodal In-Context Learning without Dedicated Data3
A Fast Maximum Mean Discrepancy based retrieval metric and a Semantics-Modality Relation-Aware retrieval metric to perform inter- and intra-dataset retrieval, respectively are proposed and found to address the limitations of existing retrieval methods in MICL without dedicated data.
- 3
- Domain Adaptive Semantic Segmentation Through Structure Enhancement3
This work proposes an effective method to adapt the segmentation network trained on synthetic images to real scenarios in an unsupervised fashion and enhances the structure information of the target images at both the feature level and the output level to improve the adaptation performance for semantic segmentation.
- A Study on the Optimization of Chain Supermarkets’ Distribution Route Based on the Quantum‐Inspired Evolutionary Algorithm3
Analysis of the current distribution situation of chain supermarkets both at home and abroad and studying the quantum-inspired evolutionary algorithm (QEA), the experiment results show that the distribution route optimized by QEA behaves better than that by GA, and QEA has stronger global search ability for both a small- scale chain supermarket and a large-scale chain supermarket.
- The Research on Controlling the Iteration of Quantum-Inspired Evolutionary Algorithms for Artificial Neural Networks3
The results show that the termination criterion proposed in this paper could control the iteration of QENN effectively and save a mass of computing time by decreasing the number of generations of Qenn.
- 2
- MSTI-Plus: Introducing Non-Sarcasm Reference Materials to Enhance Multimodal Sarcasm Target Identification2
A pluggable Semantics-aware Sarcasm Target Identification mechanism to enhance sarcasm target identification by modeling the overall semantics of sarcasm intentions via an auxiliary sample-level sarcasm recognition task is proposed.
- Progressive Multimodal Pivot Learning: Towards Semantic Discordance Understanding as Humans2
By introducing the multimodal pivot learning paradigm for multimodal recognition, the proposed PMPL approach can alleviate the negative effect of semantic discordance caused by the cross-modal information exchange mechanism of existing multimodal recognition models.
- Rethinking the Effect of Uninformative Class Name in Prompt Learning2
Information Disengagement based Associative Prompt Learning can effectively alleviate the phenomenon of learnable context overfitting to base classes, meanwhile, learning more informative semantic representation of base classes by fine-tuning the class name embedding, leading to improved performance on both base and novel classes.
- 2
- Improving Target Discriminability for Unsupervised Domain Adaptation2
This paper proposes to improve the classifier’s discriminative ability on the target domain through regularizing the entropies of the softmax predictions on thetarget data to lead to significant performance improvement for unsupervised domain adaptation.
- 1
- Causally-Aware Unsupervised Feature Selection Learning1
A novel unsupervised feature selection method, called Causally-Aware UnSupErvised Feature Selection learning (CAUSE-FS), which introduces a novel causal regularizer that reweights samples to balance the confounding distribution of each treatment feature, thus achieving causal feature selection.
- 1
- Damage Analysis via Bidirectional Multi-Task Cascaded Multimodal Fusion1
The Bidirectional Multi-task Cascaded multimodal Fusion (BiMCF) approach towards joint multimodal damage analysis is proposed to separately integrate effective visual and text information for each task, considering that different tasks attend to different information.
- Regression-based multisource conditional domain adaptation for policy outcome prediction1
An adversarial learning algorithm titled the conditional unsupervised multisource domain adversarial network (CUMDAN) is developed to address unsupervised domain adaptation in regression for policy outcome prediction.
- 1
- 1
- 1
- Learning Cross-domain Semantic-Visual Relation for Transductive Zero-Shot Learning1
This work proposes a novel approach dubbed Transferrable Semantic-Visual Relation (TSVR) to facilitate the cross-category transfer in transductive ZSL to use a domain-specific batch normalization component to reduce the domain discrepancy of semantic-visual pairs.
- Convolutional Neural Networks for Text Classification with Multi-size Convolution and Multi-type Pooling1
Experiments show that the CNN with multi-convolution and multi-type pooling (CNN-MCMP) obtains better performance on text classification compared with both the shallow machine-learning models and other CNN architectures.
- 1
- Anomaly Detection for Categorical Observations Using Latent Gaussian Process1
Categorical latent Gaussian process is adopted to capture the probabilistic distributions of conventional categorical samples and can effectively detect anomalous categorical observations and achieve better detection performance compared with other anomaly detection techniques.
- –
- Kernel Alignment-based Multi-view Unsupervised Feature Selection with Sample-level Adaptive Graph Learning–
A Kernel Alignment-based multi-view unsupervised FeatUre selection with Sample-level adaptive graph lEarning method (KAFUSE) to address feature selection issues and demonstrates the superiority of KAFUSE over state-of-the-art methods.
- –
- International Workshop on Trustworthy Multimodal Learning for Social Media Analysis–
This workshop emphasizes two themes: (i) multimodal social media analysis with LMMs, focusing on multimodal fusion and information alignment; and (ii) performance and safety evaluation for real-world deployment, covering instruction following and generation quality as well as risks such as hallucinations and ''jailbreak'' attacks.
- Regularizing modality contribution drift in multimodal continual learning–
Continual Modality Contribution Drift Regularization (CMCDR) is proposed to preserve the modality contribution profiles of previously learned tasks in both replay-based and replay-free settings to mitigate forgetting in MMCL.
- –
- –
- Cross-view Joint Learning for Mixed-Missing Multi-view Unsupervised Feature Selection–
The imputation of both missing views and variables are integrated into a feature selection model based on nonnegative orthogonal matrix factorization, enabling the joint learning of feature selection and adaptive data imputation and demonstrates that CLIM-FS outperforms state-of-the-art methods.
- Structure-aware Hybrid-order Similarity Learning for Multi-view Unsupervised Feature Selection–
This paper proposes a novel MUFS method, termed Structure-aware Hybrid-order sImilarity learner for multi-viEw unsupervised Feature Selection (SHINE-FS), which constructs a hybrid-order similarity graph that captures both local and global structures, thereby revealing the intrinsic data structure to enhance feature selection.
- –
- Dataset Awareness is not Enough: Implementing Sample-level Tail Encouragement in Long-tailed Self-supervised Learning–
Adaptive Paradigm Synergy (APS), a cross-paradigm objective that seeks to unify the strengths of both paradigms, is introduced, revealing the potential for deeper integration between supervised and self-supervised learning, paving the way for robust models that handle real-world class imbalance.
- –
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.