Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks170 from public data
- 182
- 162
- 128
- Optimized Cartesian K-Means122
This paper proposes a novel approach, named optimized cartesian K-means (ock-mean), to better encode the data points for more accurate approximate nearest neighbor search, which can provide more flexibility and lower distortion errors than traditional methods.
- Fast Scalable Supervised Hashing92
A novel supervised hashing method, called Fast Scalable Supervised Hashing (FSSH), which circumvents the use of the large similarity matrix by introducing a pre-computed intermediate term whose size is independent with the size of training data.
- GP-BPR91
This work presents a personalized compatibility modeling scheme GP-BPR, comprising of two essential components: general compatibility modeling and personal preference modeling, which characterize the item-item and user-item interactions, respectively.
- 86
- 82
- High-order nonlocal Hashing for unsupervised cross-modal retrieval79
A new deep hashing model, termed as High-order Nonlocal Hashing (HNH) to facilitate cross-modal retrieval with the following advantages: a high-order affinity measure that considers the multi- modal neighbourhood structures from a nonlocal perspective, thereby comprehensively capturing the similarity relationships between data items.
- 78
- 78
- Label Embedding Online Hashing for Cross-Modal Retrieval77
A novel supervised online cross-modal hashing method, i.e., Label EMbedding ONline hashing, LEMON, which builds a label embedding framework including label similarity preserving and label reconstructing, which may generate discriminative binary codes and reduce the computational complexity.
- 76
- Cross-Layer and Cross-Sample Feature Optimization Network for Few-Shot Fine-Grained Image Classification74
A novel Cross-layer and Cross-sample feature optimization Network for FS-FGIC, C2-Net for short, which outperforms other state-of-the-art methods by a significant margin in most cases.
- Routing Micro-videos via A Temporal Graph-guided Recommendation System71
This work designs a novel graph-based sequential network to simultaneously model users' dynamic and diverse interest and introduces users' multi-level interest into the recommendation model via a user matrix that is able to learn the enhanced representation of users' interest.
- SDMCH: Supervised Discrete Manifold-Embedded Cross-Modal Hashing69
A novel cross-modal hashing method, named Supervised Discrete Manifold-Embedded Cross-Modal Hashing (SDMCH), which can not only exploit the non-linear manifold structure of data and construct the correlation among heterogeneous multiple modalities, but also fully utilize the semantic information.
- Multi-modal Preference Modeling for Product Search68
This work proposes a multi-modal personalized product search method, which aims to search products which not only are relevant to the submitted textual query, but also match the user preferences from both textual and visual modalities.
- Chat More: Deepening and Widening the Chatting Topic via A Deep Model63
A novel deep scheme consisting of three channels, namely global, wide, and deep, which trains a Multi-layer Perceptron model to select some keywords for an in-depth discussion to generate desired responses in open-domain multi-turn dialog systems.
- DELTA: A deep dual-stream network for multi-label image classification62
A novel deep Dual-stream nEtwork for the muLTi-lAbel image classification task, DELTA, which significantly outperforms several state-of-the-art methods and can automatically locate the key image patterns that trigger the labels.
- Effective neural algorithms for the traveling salesman problem62
Two new neural algorithms are presented that improve the first one by adaptively changing the neural network and thus the optimization function and the second one is a neural implementation of the Karp and Steele algorithm based on a generalized neural network.
- SCRATCH59
This paper presents a novel supervised cross-modal hashing method, SCRATCH, which leverages the collective matrix factorization on the kernelized features and the semantic embedding with labels to find a latent semantic space to preserve the intra- and inter-modality similarities.
- User Attention-guided Multimodal Dialog Systems56
Experimental results demonstrate that the proposed model outperforms the existing state-of-the-art methods by integrating the multimodal utterances and encoding the visual features based on the users' attribute-level attention.
- Supervised Hierarchical Deep Hashing for Cross-Modal Retrieval55
This paper proposes an effective cross-modal hashing method, named Supervised Hierarchical Deep Cross- modal Hashing, SHDCH for short, to learn hash codes by explicitly delving into the hierarchical labels by directly delvinginto the hash-code learning.
- 54
- Discrete Multi-view Hashing for Effective Image Retrieval54
A novel hashing method, i.e., Discrete Multi-view Hashing (DMVH), which can work on multi-view data directly and make full use of rich information in multi-View data, and a novel approach to construct similarity matrix, which can not only preserve local similarity structure, but also keep semantic similarity between data points.
- Semi-Relaxation Supervised Hashing for Cross-Modal Retrieval54
This paper proposes a novel supervised cross-modal hashing method---Semi-Relaxation Supervised Hashing (SRSH), which can learn the hash functions and the binary codes simultaneously and relaxes a part of binary constraints, instead of all of them, by introducing an intermediate representation variable.
- 53
- A Two-Step Cross-Modal Hashing by Exploiting Label Correlations and Preserving Similarity in Both Steps52
To the best of the knowledge, this paper is the first cross-modal hashing method exploiting label correlations, and also the first two-step hashing model preserving the similarity while leaning hash function.
- 47
- Scalable Supervised Discrete Hashing for Large-Scale Search47
Scalable Supervised Discrete Hashing is presented, a novel hashing method that bypasses the direct optimization on the n by n pairwise similarity matrix and adopts no relaxation optimization scheme in the learning procedure and avoids the large quantization error problem.
- Dual Deep Neural Networks Cross-Modal Hashing46
A novel tri-stage deep cross-modal hashing method – Dual Deep Neural Networks Cross-Modal Hashing, i.e., DDCMH, which employs two deep networks to generate hash codes for different modalities and is tested on several benchmark datasets, demonstrating that it outperforms both deep and shallow state-of-the-art hashing methods.
- 45
- Personalized Capsule Wardrobe Creation with Garment and User Modeling43
This work introduces a combinatorial optimization-based personalized capsule wardrobe creation framework, named PCW-DC, which jointly integrates both garment modeling and user modeling, and constructs a dataset, named bodyFashion, which consists of user-item purchase records on Amazon involving 11,784 users and 75,695 fashion items.
- Ensemble multi-instance multi-label learning approach for video annotation task43
The En-MIMLSVM approach for the video annotation task considers the class imbalance and long time training problems of most video annotation tasks, and a temporally consistent weighted multi-instance kernel is developed to take into account both the temporal consistency in video data and the significance of instances of different levels in pyramid representation.
- An efficient simulated annealing algorithm for the minimum vertex cover problem39
In this paper, an efficient simulated annealing algorithm is presented for the minimum vertex cover problem, and an acceptance function is defined for every vertex to help the algorithm in finding a near-optimal solution to a problem.
- Machine Unlearning for Image Retrieval34
This paper makes the first attempt to realize data forgetting on deep models for image retrieval by proposing a generative scrubbing (GS) method that learns a generator to craft noisy data to manipulate the model weights.
- Asymmetric Discrete Cross-Modal Hashing32
A novel cross-modal hashing algorithm, named Asymmetric Discrete Cross-Modal Hashing (ADCH), which leverages the collective matrix factorization technique to learn the common latent representations while preserving not only the cross-correlation from different modalities but also the semantic similarity.
- Dictionary Learning Based Hashing for Cross-Modal Retrieval31
DLCMH learns dictionaries and generates sparse representation for each instance, which is more suitable to be projected to latent space and outperforms or is comparable to several state-of-the-art hashing models.
- Semi-supervised multi-instance multi-label learning for video annotation task29
A semi-supervised MIML approach, SSMIML, which is able to exploit abundant unannotated videos to help improve the annotation performance, and takes label correlations into account, and enforces similar instances to share similar multi-labels.
- Ensemble approach based on conditional random field for multi-label image and video annotation29
In this paper, the En-CRF (Ensemble based on Conditional Random Field) method is proposed, in which multiple models are first trained for each keyword, then the predictions of these models and the correlations between keywords are incorporated into a conditional random field.
- 25
- 25
- Linear unsupervised hashing for ANN search in Euclidean space25
An unsupervised hashing method - Unsupervised Euclidean Hashing (USEH), which learns and generates hashing codes to preserve the Euclidan distance relationship between data and is comparable to state-of-the-art unsuper supervised hashing methods.
- 24
- Discrete particle swarm optimization based on estimation of distribution for polygonal approximation problems24
Simulation results show that the DPSO-EDA is better than previous methods such as genetic algorithm, tabu search, particle swarm optimization, and ant colony optimization.
- High-Dimensional Sparse Cross-Modal Hashing with Fine-Grained Similarity Embedding22
An efficient sparse hashing method that not only takes the high-level semantic similarity of data into consideration, but also properly exploits the low-level feature similarity into the to-be-learnt hash codes, HSCH is presented.
- Deep Recurrent Quantization for Generating Sequential Binary Codes21
This work proposes a Deep Recurrent Quantization (DRQ) architecture which can generate sequential binary codes and achieves comparable or even better performance compared with the state-of-the-art for image retrieval.
- A method to improve the transiently chaotic neural network21
With this method, the network processes by starting at rich chaotic dynamics, and reaches stable state for all neurons rapidly after the last bifurcation, which enables the network to have rich search ability at the beginning, and use less CPU time to reach a stable state.
- 20
- Bi-directional Task-Guided Network for Few-Shot Fine-Grained Image Classification20
A novel Bi-directional Task-Guided Network (BTG-Net) is proposed to tackle the cross-task challenges of FS-FGIC, and the results demonstrate that BTG-Net outperforms state-of-the-art methods comprehensively.
- 20
- 20
- TEACH: Attention-Aware Deep Cross-Modal Hashing19
A novel cross-modal hashing method, namely aTtEntion-Aware deep Cross- modal Hashing (TEACH), that could perform feature learning and hash-code learning simultaneously and can effectively highlight the useful information of data while suppressing the redundant information.
- From Question to Text19
A question-oriented feature attention (QFA) mechanism, which learns to weight different engineering features according to the given question, so that important features with respect to the specific question is emphasized accordingly.
- 19
- 19
- 19
- Semantic Embedding Guided Attention with Explicit Visual Feature Fusion for Video Captioning18
This work designs an explicit visual-feature fusion (EVF) scheme to capture the pairwise interactions between multiple visual modalities and fuse multimodal visual features of videos in an explicit way and proposes a novel attention mechanism called semantic embedding guided attention (SEGA), which cooperates with the temporal attention to generate a joint attention map.
- Lagrangian object relaxation neural network for combinatorial optimization problems17
The Lagrangian object relaxation neural network is analyzed theoretically and evaluated experimentally through simulating the TSP to show that the proposed method can find 100% valid solutions which are near-optimal solutions.
- Few-Shot Fine-Grained Image Classification with Progressively Feature Refinement and Continuous Relationship Modeling16
A novel Progressive Feature Refinement Module (PFRM) and Continuous Relationship Modeling Module (CRMM) are proposed to capture the dependencies between query samples and the corresponding class prototypes, achieving precise optimization of the distances among corresponding sample points in the feature space.
- 16
- Online Enhanced Semantic Hashing: Towards Effective and Efficient Retrieval for Streaming Multi-Modal Data15
A novel semantic-enhanced representation for data is designed, which could help handle the new coming classes and thereby construct the enhanced semantic objective function of OASIS, an efficient and effective discrete online optimization algorithm.
- Proactive Privacy-preserving Learning for Retrieval15
A data-centric Proactive Privacy-preserving Learning (PPL) algorithm for hashing based retrieval, which achieves the protection purpose by employing a generator to transfer the original data into the adversarial data with quasi-imperceptible perturbations before releasing them.
- 14
- An automatically vetting mechanism for SSL error-handling vulnerability in android hybrid Web apps14
A hybrid approach is developed, which combines both static analysis and dynamic analysis to identify vulnerable apps and designs algorithms to select the correct paths from the app entry-point to the targeted code and determine the parameters by backtracing.
- A discrete competitive Hopfield neural network for cellular channel assignment problems14
The DCHNN can always satisfy the problem constraint and therefore guarantee the feasibility of the solutions for the CAP, and permits temporary energy increases to escape from local minima by introducing stochastic dynamics.
- POLISH: Adaptive Online Cross-Modal Hashing for Class Incremental Data13
A novel supervised online cross-modal hashing method named adaPtive Online cLass-Incremental haSHing (POLISH), which generates representations for new class label from multiple angles and treats label embeddings as stable global information to produce high-quality hash codes.
- Self-Distillation Dual-Memory Online Hashing with Hash Centers for Streaming Data Retrieval13
A new online hashing method called Self-distillation Dual-memory Online Hashing with Hash Centers, which is abbreviated to SDOH-HC, which contains replay and distillation modules to overcome the problem of catastrophic forgetting.
- S3Mix: Same Category Same Semantics Mixing for Augmenting Fine-grained Images13
The Same Category Same Semantics Mixing method (S3Mix) is proposed, which limits the mixture to regions of the same category and semantics and proposes a homology loss to promote the semantic relationship between the generated positive image pairs.
- A Multi-label Least-Squares Hashing For Scalable Image Search13
A Multi-label Least-Squares Hashing (MLSH) method for multi-label data hashing, which outperforms several state-of-the-art hashing methods including supervised and unsupervised methods.
- 12
- Weakly-Supervised Online Hashing12
In order to learn high-quality hash codes, WOH exploits the weak supervision, i.e., tags, by considering the semantics of tags and removing the noise, and develops a discrete online optimization algorithm, which is efficient and scalable.
- Dictionary Learning based Supervised Discrete Hashing for Cross-Media Retrieval12
A novel cross-modal hashing method, called Dictionary Learning based Supervised Discrete Hashing (DLSDH), which learns dictionaries and generates sparse representation for every instance, which is more suitable to be projected to a latent space.
- 11
- 11
- 10
- 10
- ViT-FOD: A Vision Transformer based Fine-grained Object Discriminator10
This work proposes a novel ViT based fine-grained object discriminator for FGVC tasks, ViT-FOD for short, and introduces three novel components, i.e, Attention Patch Combination (APC), Critical Regions Filter (CRF), and Complementary Tokens Integration (CTI).
- FedCAFE: Federated Cross-Modal Hashing with Adaptive Feature Enhancement9
Federated Cross-modal Hashing with Adaptive Feature Enhancement (FedCAFE) is a federated method which could use distributed data to train existing CMH methods under the privacy protection and is endowed with a novel adaptive feature enhancement module and a new weighted aggregation strategy.
- Effective and Robust Adversarial Training Against Data and Label Corruptions9
An Effective and Robust Adversarial Training (ERAT) framework to simultaneously handle two types of corruption without prior knowledge of their specifics is developed and is expected to enable the model to cope with unpredictable perturbations in real-world data corruption.
- 9
- IDEAL: High-Order-Ensemble Adaptation Network for Learning with Noisy Labels9
A novel\underlineIDE and \underlineAL ign (IDEAL) methodology is proposed, which aims to eliminate the feature distribution shift raised by a broad spectrum of noise patterns by identifying noisy data and clean data from the dataset as two domains and minimizing their domain discrepancy in the feature space.
- 9
- Joint-Rerank: a novel method for image search reranking9
A novel image reranking framework—Joint-Rerank is presented which considers multiple modalities of images (or multiple cues) jointly as interdependent attributes of an image entity and two methods are proposed which use different approaches to measure the inter-modal similarities between two nodes.
- 9
- DATA: Multi-Disentanglement Based Contrastive Learning for Open-World Semi-Supervised Deepfake Attribution8
An innovative multi-DisentAnglement based conTrastive leArning framework, DATA, is proposed to enhance the generalization ability on novel classes for the open-world semi-supervised deepfake attribution (OSS-DFA) task and defines the concept of ‘Orthonormal Deepfake Basis’ for the first time, thereby reducing the overfitting on forgery-irrelevant information.
- Prototype-Based Layered Federated Cross-Modal Hashing8
A novel method called prototype-based layered federated cross-modal hashing is proposed, where the prototype is introduced to learn the similarity between instances and classes on server, reducing the impact of statistical heterogeneity (non-IID) on different clients.
- Diagnose Like Doctors: Weakly Supervised Fine-Grained Classification of Breast Cancer8
This work proposes a weakly supervised fine-grained medical image classification method for breast cancer diagnosis, i.e., DLD-Net, which simulates the diagnostic procedures of pathologists by multiple attention-guided cropping and dropping operations, making it have good clinical interpretability.
- Weakly-Supervised Online Hashing with Refined Pseudo Tags8
To improve the quality of weakly-supervised tags, the real-valued pseudo tag matrix is designed and learn it by exploiting the correlation between the previous and new tags, and a memory-based similarity learning is proposed which could effectively maintain the semantic correlation between old and new data.
- 8
- DGPrompt: Dual-guidance prompts generation for vision-language models7
An original Dual-Guidance Prompts Generation (DGPrompt) method that promotes alignment between visual and textual spaces while ensuring the continuous retention of general knowledge is proposed.
- 7
- 7
- Video Captioning by Learning from Global Sentence and Looking Ahead7
Extensive experiments show that LEAD outperforms some state-of-the-art methods on MSR-VTT, MSVD, and VATEX, demonstrating the effectiveness of the proposed approach in video captioning.
- An Improved Simulated Annealing Algorithm for the Maximum Independent Set Problem7
An improved simulated annealing algorithm is presented for the maximum independent set problem, in which an acceptance function is defined for every vertex to help the algorithm find a near optimal solution to a problem.
- 6
- 6
- 5
- Hierarchical Multi-label Learning for Incremental Multilingual Text Recognition5
HierArchicalMulti-label learning framework for Multilingual tExt Recognition, termed HAMMER, is proposed, designed to identify shared knowledge and provide corresponding multi-label language supervision, and a hierarchical language evaluation mechanism to predict language scores at word and character levels.
- 5
- Semantic Enhanced Video Captioning with Multi-feature Fusion5
A novel video-captioning framework, Semantic Enhanced video captioning with Multi-feature Fusion, SEMF, which optimizes the use of different types of features from three aspects, and can be trained in an end-to-end manner.
- FedVMR: A New Federated Learning Method for Video Moment Retrieval5
This work is the very first attempt to enable safe and efficient VMR training in decentralized scene, which is hoped to pave the way for further study in the related research field.
- Hashing Multi-Instance Data from Bag and Instance Level5
This paper presents two multi-instance hashing methods: Bag-level Multi-Instance Hashing (BMIH) and Instance-level multi- instance Hashing with supervised information (IMIH), and experiments show that instance-level hashing with supervised Information outperforms all proposed techniques.
- Max-margin adaptive model for complex video pattern recognition5
The proposed max-margin adaptive (MMA) model for complex video pattern recognition, which can utilize a large number of unlabeled videos to assist the model training, is proposed.
- 5
- An Algorithm Based on Hopfield Network Learning for Minimum Vertex Cover Problem5
An efficient algorithm for the minimum vertex cover problem based on Hopfield neural network leaning is presented and it is shown that the proposed algorithm is an effective algorithm in terms of the computation time and solution quality.
- 4
- Bias Mitigating Few-Shot Class-Incremental Learning4
A novel method to mitigate model bias of the FSCIL problem during training and inference processes, which includes mapping ability stimulation, separately dual-feature classification, and self-optimizing classifiers is proposed.
- A Dual-CNN Model for Multi-label Classification by Leveraging Co-occurrence Dependencies Between Labels4
A dual-CNN model for Multi-Label classification (Dual-CNN-ML), which can make full use of the dependencies of labels to enhance classification performance and outperforms several state-of-the-art methods.
- SOMH: A self-organizing map based topology preserving hashing method4
A Self-Organizing Map based hashing framework–SOMH, which cannot only keep similarity relationship, but also preserve topology of data, is proposed and experimental results show that SOMH can outperform or is comparable to several state-of-the-arts algorithms.
- 4
- A heterogenous automatic feedback semi-supervised method for image reranking4
The experimental results show that HAFSRerank is superior or highly competitive to some state-of-the-art graph-based reranking methods, and can be well interpreted by Bayesian theory, and does not require complex search models for special queries and any additional input from users.
- Joint-rerank4
Joint-Rerank is presented, which considers image multiple modalities (or multiple cues) jointly as interdependent attributes of an image entity and the relevance scores of multiple modalities are fused to rank the images.
- 4
- Nodes Deployment Mechanism Based on Energy Efficiency in Wireless Sensor Networks4
Simulation results show that two deployment mechanisms are good and useful with energy efficiency in wireless sensor networks and the optimum value for one-hop spacing for the first mechanism is derived, and the formula to calculate the optimal value is presented.
- 4
- Enhancing HOI Detection with Contextual Cues from Large Vision-Language Models3
ConCue is proposed, a novel approach that integrates contextual cue generation with feature extraction to enhance HOI detection and leads to significant performance improvements on two widely used benchmark datasets, highlighting the potential of the approach in advancing HOI detection.
- 3
- 3
- 3
- 3
- 3
- 3
- 3
- 3
- AN IMPROVED TRANSIENTLY CHAOTIC NEURAL NETWORK FOR THE MAXIMUM INDEPENDENT SET PROBLEM3
An improved transiently chaotic neural network is presented that can yield better solutions to p-random graphs than other existing algorithms and uses fewer steps to converge to stable state in comparison with the originally proposed model.
- 2
- 2
- Three-Stream Joint Network for Zero-Shot Sketch-Based Image Retrieval2
A novel Three-Stream Joint Training Network (3JOIN) is proposed for the ZS-SBIR task, which uses a teacher network to extract the implicit semantics of the samples without the aid of other semantics and transfer the learned knowledge to unseen classes.
- BRUSH: Label Reconstructing and Similarity Preserving Hashing for Cross-modal Retrieval2
This paper presents a two-step supervised discrete hashing method, i.e., laBel ReconstrUcting and Similarity preserving Hashing (BRUSH), which simplifies the procedure of the hashing function and binary codes learning, thus improving the flexibility and efficiency.
- Supervised Class Graph Preserving Hashing for Image Retrieval and Classification2
Extensive experiments on two large scale real-world image data sets show that CGPH outperforms or is comparable to state-of-the-art hashing methods in both image retrieval and classification tasks.
- 2
- A New Neural Network Algorithm for Clique Vertex-Partition Problem2
By adding a nonlinear self-feedback to the maximum neural network (MNN), a new algorithm for the clique vertexpartition problem is proposed that introduces richer and more flexible dynamics and can prevent the network from getting stuck at local minima.
- 2
- A new motion equation for the minimum vertex cover problem2
This paper presents a new motion equation with a special term of the Hopfield neural network that works much better than the original Hopfield network and a sequential ratio-2 algorithm on random graphs, and has higher convergence rate to optimal solutions for benchmark graphs.
- 1
- 1
- 1
- 1
- 1
- Domain-Aware Semantic Alignment Hashing for Large-Scale Zero-Shot Image Retrieval1
This work proposes a novel approach called Domain-Aware Semantic Alignment Zero-Shot Hashing (DSAZH), which reveals three issues that suppress performance: semantic misalignment, biased optimization, and ambiguous Hamming distance.
- SSCD: Self-Supervised Coherence Discrimination Representation Learning for Scene Text Recognition1
A Self-Supervised Coherence Discrimination (SSCD) framework that achieves synergistic collaboration between visual features and semantic context through the following designs: in the representation learning phase, it first disentangle the scene text images into different coherence levels during the decoding phase to expand the coherence space.
- High-level Codes and Fine-grained Weights for Online Multi-modal Hashing Retrieval1
A novel approach to supervised online multi-modal hashing, called High-level Codes, Fine-grained Weights (HCFW), which introduces the concept of fine-grained weights designed to facilitate the seamless fusion of complementary multi-modal data, thereby generating multi-modal weights at the instance level and enhancing the overall hashing performance.
- 1
- 1
- Improving Hashing by Leveraging Multiple Layers of Deep Networks1
Experiments on real datasets show that the performance of end-to-end deep hashing is significantly enhanced; moreover, non-deep hashing can also benefit from the proposed technique of leveraging multiple layers’ features.
- M3LH: Multi-modal Multi-label Hashing for Large Scale Data Search1
Experimental results show that M3LH outperforms or is comparable to several state-of-the-art hashing methods on three widely used data sets and learns a project matrix which can be used to generate binary codes for out- of-samples.
- 1
- A new Multi-Instance learning method for scene classification1
Experimental results showed that the proposed Multi-Instance learning framework was on average superior to some common MI algorithms in a 5-class scene classification task.
- A new feature selection method for text categorization1
A new method based on information theory, which aims to eliminate sparsely distributed features and find features useful for categorization, showed that it could not only reduce the feature dimension to hundreds but also improve the performance.
- 1
- 1
- Towards Generalizable Deepfake Detection via Real Distribution Bias Correction–
The Real Distribution Bias Correction (RDBC) framework is introduced, which consists of two key components: the Real Population Distribution Estimation module and the Distribution-Sampled Feature Whitening module, which captures the real-world properties of real samples, thereby enhancing its generalizability to unseen target domains.
- Progressively exploring and exploiting inference data to break fine-grained classification barriers–
A novel learning paradigm is proposed that enables the model to progressively learn during inference, thereby leveraging cost-free data at inference time to more accurately represent fine-grained categories and adapt to dynamic semantic changes.
- –
- Multi-granularity Interactive Attention Framework for Residual Hierarchical Pronunciation Assessment–
- Fine-Grained Augmentation and Progressive Feature Integration for Unsupervised Fine-Grained Hashing–
This article proposes Fine-Grained Augmentation and Progressive Feature Integration for unsupervised fine-grained hashing, named FAPI, and designs fine-grained feature augmentation and cross-contrastive learning modules to enhance the capture of critical discriminative details.
- –
- –
- –
- EVC-MF: End-to-end Video Captioning Network with Multi-scale Features–
An end-to-end encoder-decoder-based network for video captioning, which efficiently utilizes multi-scale visual and textual features to generate video descriptions, and yields competitive performance compared with the state-of-theart methods.
- WITHDRAWN: SCHEMA: A Discrete Cross-Modal Hashing by Preserving Multiple Similarities–
A novel cross-modal hashing method that embeds multiple types of similarity into the learning of binary codes, i.e., the high-level similarity and the low-level Similarity, SCHEMA, which consistently outperforms some state-of-the-art hashing methods by large gaps.
- Class consistent hashing for fast Web data searching–
This work proposes a novel supervised hashing method named Class Consistent Hashing (CCH), which avoids using instance pairwise semantic similarity matrix which is widely used in existing methods, and generates hash codes for every class by optimizing the least-squares style objective function.
- –
- –
- Self-training method using cluster information of unlabeled samples–
Experimental results show that introduction of data clustering to select sample is beneficial and that the performance and convergence are better and faster than the contrastive algorithm.
- A Specialized Transiently Chaotic Neural Network for the Minimum Vertex Cover Problem–
A specialized transiently chaotic neural network is proposed for the minimum vertex cover problem that can yield better solutions to random graphs than other existing algorithms for this problem, and yield 100% convergence rate to optimal solutions to some benchmark graphs.
- An Efficient Neural Algorithm for Two-layer Planarization Problem in Graph Drawing–
A neural algorithm for the two-layer planarization problem using a gradient ascent learning of the Hopfield network is presented and shows that the proposed algorithm generates much better solutions than traditional Hopfieldnetwork and simulated annealing.
- Maximum Neural Network with Nonlinear Self-Feedback and Its Application to Maximum Independent Set Problem–
By adding a nonlinear self-feedback to the maximum neural network, this paper proposes a parallel algorithm that introduces richer and more flexible nonlinear dynamics and can prevent the network from getting stuck at local minima.
- A Modified Hopfield Neural Network for the Minimum Vertex Cover Problem–
In the modified Hopfield neural network, a correction term is introduced into the motion equation, and this correction term can find optimal or near-optimal solutions for the minimum vertex cover problem in higher probability.
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 2026-10-15. 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.