Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Partha NiyogiSuggested from co-authorship
- Hong-Jiang ZhangSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works306 from public data
- Locality Preserving Projections4,741
These are linear projective maps that arise by solving a variational problem that optimally preserves the neighborhood structure of the data set by finding the optimal linear approximations to the eigenfunctions of the Laplace Beltrami operator on the manifold.
- 2,450
- Laplacian Score for Feature Selection2,330
This paper proposes a "filter" method for feature selection which is independent of any learning algorithm, based on the observation that, in many real world classification problems, data from the same class are often close to each other.
- Unsupervised feature selection for multi-cluster data1,149
Inspired from the recent developments on manifold learning and L1-regularized models for subset selection, a new approach is proposed, called Multi-Cluster Feature Selection (MCFS), for unsupervised feature selection, which select those features such that the multi-cluster structure of the data can be best preserved.
- Fast and Accurate Matrix Completion via Truncated Nuclear Norm Regularization829
This paper proposes to achieve a better approximation to the rank of matrix by truncated nuclear norm, which is given by the nuclear norm subtracted by the sum of the largest few singular values, and develops a novel matrix completion algorithm by minimizing the Truncated Nuclear Norm.
- 805
- 758
- Locality sensitive discriminant analysis465
A novel linear algorithm for discriminant analysis, called Locality Sensitive Discriminant Analysis (LSDA), which finds a projection which maximizes the margin between data points from different classes at each local area by discovering the local manifold structure.
- 465
- 436
- Tensor Subspace Analysis407
A new algorithm called Tensor Subspace Analysis (TSA) is proposed that detects the intrinsic local geometrical structure of the tensor space by learning a lower dimensional tensor subspace and achieves better recognition rate, while being much more efficient.
- 399
- Hierarchical clustering of WWW image search results using visual, textual and link information389
A hierarchical clustering method using visual, textual and link analysis to cluster the search results into different semantic clusters of image search results is proposed.
- Hierarchical Clustering of WWW Image Search Results Using Visual389
A hierarchical clustering method using visual, textual and link analysis to cluster the search results into different semantic clusters of image search results is proposed.
- 382
- 378
- CLRNet: Cross Layer Refinement Network for Lane Detection329
This work presents Cross Layer Refinement Network (CLRNet), a novel network design which first detects lanes with high-level semantic features then performs refinement based on low-level features and introduces Line IoU loss which regresses the lane line as a whole unit to improve the localization accuracy.
- Music recommendation by unified hypergraph327
A novel music recommendation algorithm is proposed by using both multiple kinds of social media information and music acoustic-based content, and uses hypergraph to model the various objects and relations, and considers music recommendation as a ranking problem on this hypergraph.
- CrossFormer++: A Versatile Vision Transformer Hinging on Cross-Scale Attention317
A progressive group size (PGS) paradigm and an amplitude cooling layer (ACL) to alleviate the two issues that affect vision transformers’ performance, i.e., the enlarging self-attention maps and amplitude explosion are proposed.
- 257
- PI-RCNN: An Efficient Multi-Sensor 3D Object Detector with Point-Based Attentive Cont-Conv Fusion Module248
A novel fusion approach named Point-based Attentive Cont-conv Fusion(PACF) module, which fuses multi-sensor features directly on 3D points and a 3D multi-Sensor multi-task network called Pointcloud-Image RCNN(PI-RCNN as brief), which handles the image segmentation and 3D object detection tasks.
- 246
- Learning a semantic space from user's relevance feedback for image retrieval245
This work experiments with using spectral methods to infer a semantic space from user's relevance feedback, so that the system will gradually improve its retrieval performance through accumulated user interactions.
- CLIP is Also an Efficient Segmenter: A Text-Driven Approach for Weakly Supervised Semantic Segmentation240
This paper explores the potential of Contrastive Language-Image Pre-training models (CLIP) to localize different categories with only image-level labels and without further training, and proposes a novel WSSS framework called CLIP-ES.
- 219
- 219
- Speed up kernel discriminant analysis215
Spectral Regression Kernel Discriminant Analysis is presented, which casts discriminant analysis into a regression framework, which facilitates both efficient computation and the use of regularization techniques.
- Block-level link analysis203
Based on block-level link analysis, two new algorithms are proposed, Block Level PageRank and Block Level HITS, whose performances are studied extensively using web data.
- Mining hidden community in heterogeneous social networks199
Based on the observation that different relations have different importance with respect to a certain query, a new method for learning an optimal linear combination of these relations which can best meet the user's expectation is proposed.
- Learning an image manifold for retrieval196
A relevance feedback scheme which is naturally conducted only on the image manifold in question rather than the total ambient space, and an algorithmic framework to approximate the optimal mapping function by a Radial Basis Function (RBF) neural network.
- Locality sensitive semi-supervised feature selection191
A novel semi-supervised feature selection algorithm, which makes use of both labeled and unlabeled data points, which is compared with Fisher score and Laplacian score on face recognition and demonstrates the efficiency and effectiveness of the algorithm.
- 179
- Probabilistic dyadic data analysis with local and global consistency179
This work introduces a probabilistic framework for modeling both the topical and geometrical structure of the dyadic data that explicitly takes into account the local manifold structure.
- Community Mining from Multi-relational Networks178
This paper systematically analyzes the problem of mining hidden communities on heterogeneous social networks and proposes a new method for learning an optimal linear combination of these relations which can best meet the user's expectation.
- 177
- 176
- IntentGC170
This work collects abundant relationships from common user behaviors and item information, and proposes a novel framework named IntentGC to leverage both explicit preferences and heterogeneous relationships by graph convolutional networks to solve the sparsity of user-item interactions on websites.
- Locality preserving nonnegative matrix factorization170
This paper proposes a novel algorithm, called Locality Preserving Non-negative Matrix Factorization (LPNMF), and shows the encouraging results of the proposed algorithm in comparisons to the state-of-the-art algorithms on two large high-dimensional databases.
- 168
- 162
- Isometric Projection157
This paper proposes a novel linear dimensionality reduction algorithm, called Isometric Projection, which is more capable of discovering the intrinsic geometrical structure of a manifold than Principal Component Analysis (PCA).
- 151
- Robust non-negative matrix factorization143
A novel algorithm named robust nonnegative matrix factorization (RNMF) is proposed, which decomposes the non-negative data matrix as the summation of one sparse error matrix and the product of two non- negative matrices.
- Locality preserving indexing for document representation142
Experimental results show that LPI provides better representation in the sense of semantic structure, and a novel algorithm called Locality Preserving Indexing (LPI) is proposed for document indexing.
- 141
- Efficient manifold ranking for image retrieval140
The original manifold ranking algorithm is extended and a new framework named Efficient Manifold Ranking (EMR) is proposed, which significantly reduces the computational time and makes it a promising method to large scale real world retrieval problems.
- Document Summarization Based on Data Reconstruction136
This paper proposes a novel framework named Document Summarization based on Data Reconstruction (DSDR), which generates a summary which consist of those sentences that can best reconstruct the original document.
- 135
- MaCAR: Urban Traffic Light Control via Active Multi-agent Communication and Action Rectification126
This work proposes a novel Multi-agent Communication and Action Rectification (MaCAR) framework which enables active communication between agents by considering the impact of synchronous actions of agents and outperforms state-of-the-art methods on both synthetic and real-world datasets.
- Document recommendation in social tagging services119
This paper considers the problem of document recommendation using purely tagging data and proposes a novel graph-based representation learning algorithm that can generate promising recommendations and outperforms traditional recommendation algorithms.
- 117
- Video Question Answering via Hierarchical Spatio-Temporal Attention Networks117
The hierarchical spatio-temporal attention network is proposed for learning the joint representation of the dynamic video contents according to the given question and the multi-step reasoning process for open-ended video question answering is developed.
- 113
- Locality preserving clustering for image database111
This paper proposes a clustering algorithm named Locality Preserving Clustering (LPC), which shares many of the data representation properties of nonlinear spectral method, yet provides an explicit mapping function which is defined everywhere, both on training data points and testing points.
- Model Compression and Efficient Inference for Large Language Models: A Survey110
This paper investigates compression and efficient inference methods for large language models from an algorithmic perspective, and distinguishes large language models into medium models and ``real'' large models.
- Learning to Affiliate: Mutual Centralized Learning for Few-shot Classification110
This paper proposes a novel Mutual Centralized Learning (MCL) to fully affiliate these two disjoint dense features sets in a bidirectional paradigm and relates its method to learning a centrality on an affiliation network and demonstrates its capability to be plugged in existing methods by highlighting centralized local features.
- Deep Reinforcement Learning for Sponsored Search Real-time Bidding110
This paper forms a robust MDP model at hour-aggregation level of the auction data and proposes a control-by-model framework for SS-RTB, a reinforcement learning (RL) solution for handling the complex dynamic environment in sponsored search auction.
- DASA: Dissatisfaction-oriented Advertising based on Sentiment Analysis109
This paper proposes a novel advertising strategy, called Dissatisfaction-oriented Advertising based on sentiment analysis (DASA), to simultaneously improve ad relevance and user experience and designs a prototype system for product information submission for the sake of ad selection.
- GD-MAE: Generative Decoder for MAE Pre-Training on LiDAR Point Clouds108
This work proposes a Generative Decoder for MAE (GD-MAE) to automatically merges the surrounding context to restore the masked geometric knowledge in a hierarchical fusion manner and achieves comparable accuracy even with 20% of the labeled data on the Waymo dataset.
- Regularized locality preserving indexing via spectral regression107
Benefit from recent progresses on spectral graph analysis, the original LPI algorithm is cast into a regression framework which enable us to avoid eigen-decomposition of dense matrices and this makes it an efficient and effective data preprocessing method for large scale text clustering, classification and retrieval.
- 105
- Spectral regression104
By using techniques from spectral graph embedding and regression, this paper proposes a unified framework, called spectral regression, for learning an image subspace, which provides much faster computation and therefore makes the retrieval system capable of responding to the user's query more efficiently.
- Dialogue Act Recognition via CRF-Attentive Structured Network102
This paper tackles the problem of DAR from the viewpoint of extending richer Conditional Random Field (CRF) structured dependencies without abandoning end-to-end training and incorporates hierarchical semantic inference with memory mechanism on the utterance modeling at multiple levels.
- 102
- Whom to mention99
An in-depth study of mention mechanism and a recommendation scheme to solve the essential question of whom to mention in a tweet are presented and the advantage of the proposed algorithm compared with the traditional recommendation methods is demonstrated.
- Incremental semi-supervised subspace learning for image retrieval98
A semi-supervised subspace learning algorithm which incrementally learns an adaptive subspace by preserving the semantic structure of the image space, based on user interactions in a relevance feedback driven query-by-example system is proposed.
- Probabilistic question recommendation for question answering communities97
This paper adopts the Probabilistic Latent Semantic Analysis (PLSA) model for question recommendation and proposes a novel metric to evaluate the performance of the approach, and shows the experimental results show the recommendation approach is effective.
- 96
- 94
- Parallel field alignment for cross media retrieval91
A novel method for the cross media retrieval task, named Parallel Field Alignment Retrieval (PFAR), which integrates a manifold alignment framework from the perspective of vector fields, and can effectively preserve the metric of data manifolds and project their relationship into intermediate latent semantic spaces during the process of manifold alignment.
- Gaussian Mixture Model with Local Consistency91
A novel method based on manifold structure for data clustering, called Locally Consistent Gaussian Mixture Model (LCGMM), is introduced, which construct a nearest neighbor graph and adopt Kullback-Leibler Divergence as the distance measurement to regularize the objective function of GMM.
- UniPAD: A Universal Pre-Training Paradigm for Autonomous Driving90
UniPAD is presented, a novel self-supervised learning paradigm applying 3D volumetric differentiable rendering that implicitly encodes 3D space, facilitating the reconstruction of continuous 3D shape structures and the intricate appear-ance characteristics of their 2D projections.
- Expert finding for community-based question answering via ranking metric network learning88
This paper proposes a novel ranking metric network learning framework for expert finding by exploiting both users' relative quality rank to given questions and their social relations and develops a random-walk based learning method with recurrent neural networks forranking metric network embedding.
- 86
- COP: Customized Deep Model Compression via Regularized Correlation-Based Filter-Level Pruning84
A novel algorithm named COP (correlation-based pruning) is developed, which can detect the redundant filters efficiently and enable the cross-layer filter comparison through global normalization, and adds parameter-quantity and computational-cost regularization terms to the importance.
- Continuous Influence Maximization82
This paper forms the general continuous influence maximization problem, investigates the essential properties, and develops a general coordinate descent algorithm as well as the engineering techniques for practical implementation, and demonstrates that more efficient methods are feasible under specific influence models.
- PVT-SSD: Single-Stage 3D Object Detector with Point-Voxel Transformer81
A novel Point-Voxel Transformer for single-stage 3D detection (PVT-SSD) that takes advantage of voxel-based sparse convolutions for efficient feature encoding and adaptively fuses long-range contextual and local geometric information around reference points into content queries.
- Using Graph Model for Face Analysis80
It is shown that LPP provides a more general framework for subspace learning and a natural solution to the small sample issue in LDA.
- 76
- LoRA-Composer: Leveraging Low-Rank Adaptation for Multi-Concept Customization in Training-Free Diffusion Models75
LoRA-Composer is introduced, a training-free framework designed for seamlessly integrating multiple LoRAs, thereby enhancing the harmony among different concepts within generated images, and two inference techniques to accelerate inference speed without performance degradation and enhance the accuracy of the generated region, respectively.
- Heterogeneous hypergraph embedding for document recommendation75
A heterogeneous hypergraph model which fully exploits high-order relational information in tagging data and, meanwhile, customizes the influence of different types of relations is proposed and a novel heterogeneoushypergraph embedding framework is developed for document recommendation.
- Learning and inferring a semantic space from user's relevance feedback for image retrieval74
This work experiments with using spectral methods to infer a semantic space from user's relevance feedback, so the system will gradually improve its retrieval performance through accumulated user interactions.
- Tracking people in RGBD videos using deep learning and motion clues72
This work proposes to train a deep convolutional neural network, which improves tracking performance, to classify people and a motion model based on spatial and kinetic clues is combined with the network to track people in the scene.
- Statistical and computational analysis of locality preserving projection72
Different from Principal Component Analysis (PCA), LPP obtains a subspace spanned by the smallest eigenvectors of the local covariance matrix, which shows that the performance can be significantly improved in the subspace, and especially LPP works much better than PCA.
- Orthogonal locality preserving indexing72
Orthogonal LPI is insensitive to the number of dimensions, which makes it an efficient data preprocessing method for text clustering, classification, retrieval, etc, and shows that OLPI can have more locality preserving power than LPI.
- 72
- Knowing a tree from the forest72
The collaborative ensemble learning, a novel statistical learning approach to this task, applies probabilistic support vector machines (SVMs) to model each individual user's profile based on given examples, i.e. liked or disliked paintings.
- Translating Embeddings for Knowledge Graph Completion with Relation Attention Mechanism71
Extensive experiments show that the proposed novel knowledge graph embedding method named TransAt outperforms state-of-the-art methods significantly on public datasets, and the method can learn the true attention varying among relations.
- 71
- Spectral Regression for Dimensionality Reduction68
Spectral Regression casts the problem of learning an embedding function into a regression framework, which avoids eigen-decomposition of dense matrices and can be performed in supervised, unsupervised and semisupervised situation.
- 67
- Neural collapse inspired attraction–repulsion-balanced loss for imbalanced learning65
This paper proposes Attraction-Repulsion-Balanced Loss (ARB-Loss) to balance the different components of the gradients of the classifier weights, which can achieve state-of-the-art performance via only one-stage training instead of 2-stage learning like nowadays SOTA works.
- Stability and Hopf bifurcation analysis for a Lotka-Volterra predator-prey model with two delays64
Stability and Hopf bifurcation analysis for a two-species Lotka-Volterra predator-prey model with two delays and the linear stability of the positive equilibrium is investigated and hopf bIfurcation is demonstrated.
- Subspace Learning Based on Tensor Analysis64
This paper considers an image as the second order tensor in Rm1⊗Rm2 and proposes two novel tensor subspace learning algorithms called TensorPCA and TensorLDA, which explicitly take into account the relationship between column vectors of the image matrix and the relationships between the row vectors ofThe image matrix.
- Tensor space model for document analysis63
This paper proposes a novel Tensor Space Model for document analysis, which represents documents as the second order tensors, or matrices, and develops a novel indexing algorithm called Tensor Latent Semantic Indexing (TensorLSI), developed in the tensor space.
- 62
- Clustering and searching WWW images using link and page layout analysis61
iFind is described, a system for clustering and searching WWW images, which is less sensitive to noisy links than previous methods like PageRank, HITS, and PicASHOW, and hence the image graph can better reflect the semantic relationship between images.
- Community-Based Question Answering via Asymmetric Multi-Faceted Ranking Network Learning59
This paper proposes a novel asymmetric multi-faceted ranking network learning framework for community-based question answering by jointly exploiting the deep semantic relevance between question-answer pairs and the answerers' authority to the given question.
- 59
- 58
- 56
- 56
- 56
- SimulSLT54
SimulSLT, the first end-to-end simultaneous sign language translation model, which can translate sign language videos into target text concurrently, is proposed, composed of a text decoder, a boundary predictor, and a masked encoder.
- Microblog Sentiment Classification via Recurrent Random Walk Network Learning54
This paper proposes a novel recurrent random walk network learning framework for the microblog sentiment classification problem by exploiting both users’ posted tweets and their social relations in microblogs, and introduces the deep recurrent neural networks with random-walk layer for heterogeneous MSC network embedding.
- A low rank structural large margin method for cross-modal ranking52
A general cross-modal ranking algorithm to optimize the listwise ranking loss with a low rank embedding, which is called Latent Semantic Cross-Modal Ranking (LSCMR) and shows significant improvements over the state-of-the-art methods.
- 51
- Multi-Turn Video Question Answering via Multi-Stream Hierarchical Attention Context Network50
This paper proposes the hierarchical attention context network for context-aware question understanding by modeling the hierarchically sequential conversation context structure and develops the multi-stream spatio-temporal attention network for learning the joint representation of the dynamic video contents and context- aware question embedding.
- Learning with Tensor Representation47
This paper proposes two novel algorithms called Support Tensor Machines (STM) and Tensor Least Square (TLS), which operate in the tesnor space and aims at finding a minimum residual sum-of-squares classifier in the tensor space.
- Video Question Answering via Hierarchical Dual-Level Attention Network Learning46
This paper develops the hierarchical duallevel attention networks to learn the question-aware video representations with word-level and question-level attention mechanisms, and devise thequestion-level fusion attention mechanism for the proposed networks to learning the questionaware joint video representation.
- Iterative Multi-View Hashing for Cross Media Indexing46
This paper studies the cross media indexing problem by learning the discriminative hashing functions to map the multi-view datum into a shared hamming space and proposes a novel hashing algorithm called Iterative Multi-View Hashing (IMVH), which not only greatly improves the retrieval accuracy but also performs strong robustness.
- Motion-Aware Memory Network for Fast Video Salient Object Detection45
A space-time memory (STM)-based network that employs a standard encoder–decoder architecture that extracts high-level temporal features from the current frame and its adjacent frames, which is more efficient and practical than methods reliant on optical flow.
- Why Do We Click45
This work devise the local impression modeling module to simultaneously attend to decomposed details in the impression when understanding the semantic meaning of news title, which could explicitly get close to the process of users reading news.
- Image clustering with tensor representation45
This paper represents the images as points in the tensor space Rn1 mathcal Rn2 which is a tensor product of two vector spaces and proposes a novel image representation and clustering algorithm which explicitly considers the manifold structure of the Tensor space.
- Improving face recognition with domain adaptation43
This paper crawls a large face dataset whose distribution is different from LFW and shows the improvement of LFW accuracy with a simple domain adaptation technique, which is the first time that domain adaptation is applied in the unconstrained face recognition problem with a million scale dataset.
- Image representation using Laplacian regularized nonnegative tensor factorization43
This paper proposes a novel Laplacian regularized nonnegative tensor factorization (LRNTF) method for image representation and clustering and demonstrates that LRNTF achieves better image clustering performance, while being more insensitive to noise.
- 42
- 42
- Local Feature Discriminant Projection42
This paper introduces the Differential Scatter Discriminant Criterion (DSDC) to the subspace learning of local feature descriptors which avoids the matrix singularity problem and proposes a generalized orthogonalization method to impose on projections, leading to a more compact and less redundant subspace.
- Discourse Marker Augmented Network with Reinforcement Learning for Natural Language Inference41
This paper proposes to transfer knowledge from some important discourse markers to augment the quality of the NLI model by using reinforcement learning to optimize a new objective function with a reward defined by the property of theNLI datasets to make full use of the labels information.
- Laplacian optimal design for image retrieval40
A novel active learning algorithm, called Laplacian Optimal Design (LOD), is proposed, based on aregression model which minimizes the least square error on the measured (or, labeled) images and simultaneously preserves the local geometrical structure of the image space.
- DMN4: Few-Shot Learning via Discriminative Mutual Nearest Neighbor Neural Network39
It is argued that a Mutual Nearest Neighbor (MNN) relation should be established to explicitly select the query descriptors that are most relevant to each task and discard less relevant ones from aggregative clutters in FSL.
- Graph embedding with constraints39
This paper proposes a novel dimensionality reduction algorithm, called Constrained Graph Embedding, which considers the label information as additional constraints, and constrain the space of the solutions that are explored only to contain embedding results that are consistent with the labels.
- Query rewriting using active learning for sponsored search39
This study uses an active learning algorithm called Transductive Experimental Design to select the most informative samples to train the query rewriting relevance model and shows that this approach significantly improves model accuracy and rewriting relevance.
- Online robust principal component analysis via truncated nuclear norm regularization37
A novel online RPCA algorithm is proposed by adopting a recently proposed truncated nuclear norm as a tighter approximation of low rank constraint and decompose the objective function as a summation of sample-wise cost.
- Tag-oriented document summarization36
Experimental results show the tag-oriented summarization has a significant improvement over those not using tags, and a new tag ranking algorithm named EigenTag is proposed in this paper to reduce noise in tags.
- Online Kernel Learning with a Near Optimal Sparsity Bound35
This work focuses on Online Sparse Kernel Learning that aims to online learn a kernel classifier with a bounded number of support vectors and shows promising performance of the proposed algorithm compared to the state-of-the-art algorithms for online sparse kernel learning.
- 33
- 33
- PriorLane: A Prior Knowledge Enhanced Lane Detection Approach Based on Transformer32
A novel and general framework called PriorLane is proposed, which is used to enhance the segmentation performance of the fully vision transformer by introducing the low-cost local prior knowledge.
- 32
- Learning search tasks in queries and web pages via graph regularization32
A novel graph-based regularization algorithm is designed for search task prediction by leveraging the graph to simultaneously classify queries and web pages into the popular search tasks by exploiting their content together with click-through logs.
- 31
- 31
- Accelerated singular value thresholding for matrix completion30
An Accelerated Singular Value Thresholding (ASVT) algorithm is proposed which improves the convergence rate from O(1/N) for SVT to O( 1/N2), where N is the number of iterations during optimization.
- ES-Net: Erasing Salient Parts to Learn More in Re-Identification29
A novel network, Erasing-Salient Net (ES-Net), to learn comprehensive features by erasing the salient areas in an image by using a trainable pooling layer P-pooling that generalizes global max and global average pooling.
- An Agent Organization Theory and Its Application to Wide Area Backup Protection29
A novel formal representation of agent organization is proposed which gives a theoretical basis for multi-Agent system and the test results illustrate the rationality of the model, and provide a reference for the engineering development of WABP systems.
- WeakM3D: Towards Weakly Supervised Monocular 3D Object Detection28
The weakly supervised monocular 3D detection method is explored, which first detects 2D boxes on the image and adopts corresponding RoI LiDAR points as the weak supervision, then adopts a network to predict 3D boxes which can tightly align with associated RoILiDar points.
- 28
- Parallel vector field embedding28
A novel local isometry based dimensionality reduction method from the perspective of vector fields, which is called parallel vector field embedding (PFE), which can precisely recover the manifold if it is isometric to a connected open subset of Euclidean space.
- 27
- 27
- Salient Object Ranking with Position-Preserved Attention27
This paper proposes the first end-to-end framework of the SOR task and solves it in a multi-task learning fashion, and outperforms the former state-of-the-art method significantly.
- Image retrieval based on incremental subspace learning27
This paper proposes a semi-supervised subspace learning algorithm for image retrieval based on LPP which can incorporate user's relevance feedbacks and can ultimately obtain a semantic subspace in which different semantic classes can be best separated and the retrieval performance can be enhanced.
- X-View: Non-Egocentric Multi-View 3D Object Detector26
This paper proposes a novel multi-view-based 3D detection method, named X-view, which breaks through the traditional limitation about the perspective view whose original point must be consistent with the 3D Cartesian coordinate and shows that X-View obtains consistent improvements when combined with mainstream state-of-the-art 3D methods.
- A-Optimal Projection for Image Representation26
This paper proposes a novel dimensionality reduction algorithm called A-Optimal Projection (AOP), based on a linear regression model, which finds the optimal basis functions so that the expected prediction error of the regression model can be minimized if the new representations are used for training the model.
- 26
- Advertising keyword generation using active learning26
This paper proposes an efficient relevance feedback based interactive model for keyword generation in sponsored search advertising that improves the relevance of generated terms significantly with little user effort required.
- Non-negative matrix factorization with sinkhorn distance25
This paper adopts an approximation of Earth Mover's Distance as metric and add a graph regularized term based on EMD to NMF and proposes an efficient multiplicative iteration algorithm to solve it.
- 25
- 25
- 25
- 24
- Global existence of periodic solutions in a six-neuron BAM neural network model with discrete delays24
A six-neuron BAM neural network model with discrete delays is considered and a set of sufficient conditions for the system to have multiple periodic solutions are derived when the sum of delays is sufficiently large.
- 23
- 23
- Discriminative concept factorization for data representation23
An iterative algorithm is developed to solve the optimization problem through alternative convex programming and results in a concept factorization adapted to the classification task and a classifier built on the low-dimensional representations.
- Sparse projections over graph23
This paper presents a novel technique for learning a sparse projection over graphs based on a graph embedding model, which encodes the discriminating and geometrical structure in terms of the data affinity and applies regularized regression to learn a set of sparse basis functions.
- A geometric viewpoint of manifold learning22
This paper categorizes manifold learning algorithms by three main groups: Laplacian-based, Hessian- based, and parallel field-based methods and shows the connection and difference between these three groups on their continuous and discrete counterparts.
- Convex experimental design using manifold structure for image retrieval22
This paper proposes a novel active learning algorithm, called Convex Laplacian Regularized Ioptimal Design (CLapRID), based on a regression model which minimizes the least square error on the labeled images and simultaneously preserves the intrinsic geometrical structure of the image space.
- News article extraction with template-independent wrapper22
This paper formalizes news extraction as a machine learning problem and learns a template-independent wrapper using a very small number of labeled news pages from a single site.
- Does Network Width Really Help Adversarial Robustness21
It is shown that the model robustness is closely related to both natural accuracy and perturbation stability, a new metric proposed in this paper to characterize the model's stability under adversarial perturbations, and that to fully unleash the power of wide model architecture, practitioners should adopt a larger regularization parameter for training wider networks.
- 21
- 21
- 21
- Image retrieval using nonlinear manifold embedding21
KMMP is a novel semi-supervised learning method for dimensionality reduction designed for discovering the local manifold structure and after projecting the images into a lower dimensional subspace, KMMP significantly improves the performance of image retrieval.
- Efficient Online Learning for Large-Scale Sparse Kernel Logistic Regression20
To improve the sparsity of the KLR classifier, two conservative online learning algorithms are proposed that update the classifier in a stochastic manner and generate sparse solutions and achieve similar classification accuracy but with significantly shorter training time.
- Attentional Image Retweet Modeling via Multi-Faceted Ranking Network Learning20
A novel attentional multi-faceted ranking network learning framework with multi-modal neural networks is developed for the proposed heterogenous IRM network to learn the joint image tweet representations and user preference representations for prediction task.
- 20
- Sparse Learning for Stochastic Composite Optimization20
A simple but effective stochastic optimization scheme that adds a novel sparse online-to-batch conversion to the traditional SCO algorithms for sparse learning and can find a solution with better sparse patterns without affecting the convergence rate.
- SimulLR: Simultaneous Lip Reading Transducer with Attention-Guided Adaptive Memory19
A novel attention-guided adaptive memory is devised to organize semantic information of history segments and enhance the visual representations with acceptable computation-aware latency in the task of simultaneous lip reading.
- Sparse Learning with Stochastic Composite Optimization19
This paper proposes a simple yet effective two-phase Stochastic Composite Optimization scheme by adding a novel powerful sparse online-to-batch conversion to the general Stochastics Optimization algorithms.
- Regularized regression on image manifold for retrieval19
This paper proposes a novel approach for image retrieval on manifold based on a regularized linear regression framework where the local manifold structure and user-provided relevance feedbacks are incorporated into the image retrieval system through a Locality Preserving Regularizer.
- Unsupervised face-name association via commute distance18
A novel framework named face- name association via commute distance (FACD), which judges face-name and face-null assignments under a unified framework via commutedistance (CD) algorithm, and a novel anchor-based commute Distance (ACD) algorithm whose main idea is using the anchor point representation structure to accelerate the eigen-decomposition of the adjacency matrix of a graph.
- Discriminative codeword selection for image representation18
The central idea of the proposed algorithm is to select those codewords so that the cluster structure of the image database can be best respected, and a multi-output linear function is fitted to model the relationship between the data matrix after codeword selection and the indicator matrix.
- VLAD-VSA: Cross-Domain Face Presentation Attack Detection with Vocabulary Separation and Adaptation16
The VLAD aggregation method is adopted to quantize local features with visual vocabulary locally partitioning the feature space, and hence preserve the local discriminability, and the proposed vocabulary separation and adaptation method imitates the maximization step of the k-means algorithm in the end-to-end training.
- 16
- 16
- Link Prediction via Ranking Metric Dual-Level Attention Network Learning16
This paper proposes a novel proximity ranking metric attention network learning framework by jointly exploiting both node-level and path-level attention proximity of the endpoints to their betweenness paths for learning the discriminative feature representation for link prediction.
- 16
- Tangent space learning and generalization16
Inspired from the statistics on Grassmann manifold, this work uses intrinsic sample total variance to measure the variation of estimated tangent spaces at a single point, and thus, the generalization problem can be solved by estimating the intrinsic sample mean on Grassman manifold.
- 15
- Partial Multi-Modal Sparse Coding via Adaptive Similarity Structure Regularization15
This paper proposes a partial multi-modal sparse coding framework termed Adaptive Partial Multi-Modal Similarity Structure Regularization for Sparse Coding (AdaPM2SC), which preserves the similarity structure within the same modality and between different modalities.
- Large scale multi-class classification with truncated nuclear norm regularization15
A new scalable large scale multi-class classification algorithm is proposed by using the recently proposed truncated nuclear norm as a better surrogate of the rank operator of matrices along with multinomial logisitic loss to solve the non-convex and non-smooth optimization problem.
- 15
- 15
- Semi-Supervised Regression using Spectral Techniques15
This paper proposes a novel method, called Spectral Regression (SR), which first compute a set of responses for each sample which respects both the label information and geometrical structure, and is particularly designed for multi-class problem.
- 14
- Open-Ended Long-Form Video Question Answering via Hierarchical Convolutional Self-Attention Networks14
A hierarchical convolutional self-attention encoder to efficiently model long-form video contents, which builds the hierarchical structure for video sequences and captures question-aware long-range dependencies from video context and a multi-scale attentive decoder to incorporate multi-layer video representations for answer generation.
- Regularized query classification using search click information14
This paper presents experimental evidence suggesting that the regularized regression algorithm is able to use search click information effectively for query classification, and selects the one which best preserves the semantic relationship between queries.
- 13
- Continuous Influence Maximization13
This article forms the general continuous influence maximization problem, investigates the essential properties, and develops a general coordinate descent algorithmic framework as well as the engineering techniques for practical implementation.
- Accelerated Sparse Linear Regression via Random Projection13
Theoretical analysis shows that the proposed accelerated numerical method solves the sparse linear regression problem via a proximal gradient method with a homotopy strategy to generate sparse intermediate solutions and achieves a global geometric convergence rate.
- 13
- Semi-supervised Support Vector Learning for Face Recognition13
This paper introduces a novel semi-supervised learning method for face recognition based on Support Vector Machine, which states that, if two data points are close to each other, they tend to share the same label.
- Parallel implementation and optimization of high definition video real-time dehazing12
This paper proposes a parallel implementation and optimization method for the real-time dehazing of the high definition videos based on a single image haze removal algorithm that takes full advantage of the modern CPU+GPU architecture, which increases the parallelism of the algorithm, and greatly reduces the computational complexity and the execution time.
- 11
- Progressive Transfer Learning for Person Re-identification11
This paper introduces a novel network structure called Batch-related Convolutional Cell (BConv-Cell), which progressively collects the global information of the dataset into a latent state and uses this latent state to rectify the extracted feature.
- Multi-label active learning based on submodular functions11
A query strategy by constructing a submodular function for the selected instance-label pairs, which can measure and combine the informativeness and representativeness is proposed, which outperforms several state-of-the-art multi-label active learning methods.
- A Class of Manifold Regularized Multiplicative Update Algorithms for Image Clustering11
A class of novel graph regularized NMF algorithms is developed, which results in a class of extended multiplicative update algorithms with manifold structure regularization, which shows that the proposed algorithms can achieve state-of-the-art performance in applications of image clustering.
- 11
- 11
- A discrete analogue of Lyapunov-type inequalities for nonlinear difference systems11
This paper establishes several new Lyapunov-type inequalities for the nonlinear difference system when the end-points are not necessarily usual zeros, but rather, generalized zeros.
- 11
- 10
- 10
- 10
- 10
- Atom Decomposition with Adaptive Basis Selection Strategy for Matrix Completion9
A new efficient and scalable atom decomposition algorithm for matrix completion called Adaptive Basis Selection Strategy (ABSS), which outperforms the state-of-the-art SVD-based matrix completion algorithms by similar or better reconstruction accuracy with enormous advantages on efficiency.
- Oriented boundary padding for iterative and oriented fringe pattern denoising techniques9
An oriented padding method is proposed for iterative and oriented fringe pattern denoising techniques that require partial derivative estimations and helps to reduce the error.
- Multi-task Vector Field Learning.9
This paper develops multi-task vector field learning (MTVFL) which learns the predictor functions and the vector fields simultaneously and formalizes the idea in a regularization framework and provides a convex relaxation method to solve the original non-convex problem.
- 8
- 8
- 8
- 8
- Error-correcting output hashing in fast similarity search8
This paper proposes a new method called Error-correcting Output Hashing (ECOH) which meets all the above three requirements and experimental results on real world data sets demonstrate the effectiveness and efficiency of the proposed approach.
- Random Projection with Filtering for Nearly Duplicate Search7
The key idea is to introduce a filtering procedure within the search algorithm, based on the compressed sensing theory, that effectively removes the false positive answers.
- 7
- Decouple co-adaptation: Classifier randomization for person re-identification7
This paper proposes a very simple and effective training method, called DeAda, to construct a series of weak classifiers during training by randomization of parameters, so that optimization on the feature extractor could be strengthened in the training stage.
- 7
- 7
- 7
- 7
- Any-to-Bokeh: Arbitrary-Subject Video Refocusing with Video Diffusion Model6
This work proposes a novel one-step video bokeh framework that converts arbitrary input videos into temporally coherent, depth-aware bokeh effects and leverages a multi-plane image (MPI) representation constructed through a progressively widening depth.
- Local Conditional Controlling for Text-to-Image Diffusion Models6
This paper proposes Regional Discriminate Loss to update the noised latents, aiming at enhanced object generation in non-control regions, and the proposed Focused Token Response suppresses weaker attention scores which lack the strongest response to enhance object distinction and reduce duplication.
- Spherical Knowledge Distillation.6
Spherical Knowledge Distillation is proposed, which project the teacher and the student's logits into a unit sphere, and then it is argued that the norm of logits is actually interference, which damages the efficiency in the transfer process.
- 6
- G-Optimal Feature Selection with Laplacian regularization6
A novel feature selection approach called Laplacian G-Optimal Feature Selection (LapGOFS) is proposed which minimizes the maximum variance of the predicted value of the regression model.
- 6
- Discriminator-Free Direct Preference Optimization for Video Diffusion5
This work proposes a discriminator-free video DPO framework that eliminates the need for costly synthetic video comparisons, provides unambiguous quality signals, and enables unlimited training data expansion through simple editing operations.
- 5
- 5
- 5
- Unsupervised document summarization from data reconstruction perspective5
This paper proposes a novel unsupervised framework named Document Summarization based on Data Reconstruction (DSDR), which generates a summary which consist of those sentences that can best reconstruct the original document.
- Salient Object Detection via Fast Iterative Truncated Nuclear Norm Recovery5
This paper proposes a novel salient object detection method called Fast Iterative Truncated Nuclear Norm Recovery (FIT) to detect salient objects and demonstrates the efficiency of the proposed algorithm compared with the state-of-the-art.
- Parallel field ranking5
This paper proposes a novel ranking algorithm on the data manifolds, called Parallel Field Ranking, which tries to learn a ranking function and a vector field simultaneously, and requires the vector field to be close to the gradient field of the ranking function, and the vector field to be as parallel as possible.
- 5
- On Lyapunov-type inequalities for nonlinear dynamic systems on time scales5
By using elementary analysis, several new Lyapunov type inequalities are established for the following nonlinear dynamic system on an arbitrary time scale, when the end-points are not necessarily usual zeros, but rather, generalized zeros.
- Approximately harmonic projection: Theoretical analysis and an algorithm5
This paper provides theoretical analysis on both the functional and the constraint, which shows the affine hulls of the manifold and the connected components are essential to linear manifold learning problem.
- Few-Shot Domain Adaptation for Charge Prediction on Unprofessional Descriptions4
This paper proposes a novel few-shot domain adaptation (FSDA) method named Disentangled Legal Content for Charge Prediction (DLCCP), and contributes the first publicly available non-PLLS dataset named NCCP for developing layperson-friendly charge prediction models.
- Multi-scale fusion transformer based weakly supervised hashing learning for instance retrieval4
A hierarchical with a spatial pyramidal structure for weakly supervised multi-instance hash learning that merges the advantages of local and multi-scale perception on CNN with the global field of view on Transformer.
- CCL4Rec: Contrast over Contrastive Learning for Micro-video Recommendation4
The Contrast over Contrastive Learning framework for training recommender models is proposed, named CCL4Rec, which models the nuances of different augmented views by further contrasting augmented positives/negatives with adaptive pulling/pushing strengths, i.e., the contrast over (vanilla) contrastive learning.
- 4
- 4
- CoHomo: A cluster-attribute correlation aware graph clustering framework4
A novel attributed graph clustering framework called C o H omo is proposed to detect clusters that show both high structural cohesiveness and value homogeneity on some of the attributes and formalize this problem as a bi-objective optimization problem and design an efficient heuristic optimizing method.
- Split-Net: Improving face recognition in one forwarding operation4
This paper proposes a unified end-to-end framework, named as Split-Net, which splits selective intermediate feature maps into several branches instead of cropping on original images, and can effectively improve the accuracy of face recognition with less computation increased.
- Robust Asymmetric Bayesian Adaptive Matrix Factorization4
This paper proposes robust Asymmetric Laplace Adaptive Matrix Factorization model(ALAMF) under bayesian matrix factorization framework and assumption of Laplace distribution makes the model more robust and the asymmetric attribute makes the authors' model more flexible and adaptable to real-world noise.
- 4
- K-way min-max cut for image clustering and junk images filtering from Google images4
This work applies K-way min-max cut to cluster images returned by Google into multiple clusters based on the mixture of feature kernels, with kernel weights being determined automatically instead of hard fix to filter junk images.
- Image Retrieval Using Dimensionality Reduction4
This paper investigates the use of Laplacian Eigenmap (LE) for image representation and retrieval and compares the PCA and LE based image representations in the context of image retrieval.
- 3
- 3
- 3
- Task-Agnostic Self-Distillation for Few-Shot Action Recognition3
This work proposes a novel meta-distillation framework for few-shot action recognition that learns the task-agnostic metric features and generalizes them to different tasks and demonstrates its superior performance compared to existing state-of-the-art methods.
- Towards In-Distribution Compatible Out-of-Distribution Detection3
This work identifies three causes of the in-distribution incompatibility: contradictory gradient, false likelihood, and distribution shift and proposes a new out-of-dist distribution detection method by adapting both the top-design of deep models and the loss function.
- PointCSE: Context-sensitive encoders for efficient 3D object detection from point cloud3
The proposed encoder called PointCSE outperforms prior real-time encoders by a large margin with 1.5x memory reduction and achieves state-of-the-art performance with 49 FPS inference speed.
- 3
- Multi-Query Parallel Field Ranking for image retrieval3
This paper proposes a novel approach called Multi-Query Parallel Field Ranking (MQPFR) which finds an optimal ranking function whose gradient field is as parallel as possible and achieves the highest value at the multiple queries simultaneously.
- Active Learning Based on Local Representation.3
This paper proposes a novel framework named Active Learning via Neighborhood Reconstruction (ALNR) by taking into account the locality information directly during the selection for the linear reconstruction of target point by developing an efficient two-stage iterative procedure to solve the final optimizationproblem.
- Large-Scale Storage Location Assignment via Hierarchical Reinforcement Learning: A Rank and Assign Approach2
A new buffer-wise framework based on hierarchical reinforcement learning for EC-SLAP, aimed at optimizing container turnover efficiency, and a novel immediate reward function to enhance learning efficiency and accelerate convergence are proposed.
- Self-Supervised Direct Preference Optimization for Text-to-Image Diffusion Models2
Self-supervised direct preference optimization is introduced, a new paradigm that removes the need for any pre-generated images or manual ranking and eliminates costly data collection and annotation while remaining plug-and-play for any text-to-image diffusion method.
- Unsupervised Entity Alignment Based on Personalized Discriminative Rooted Tree2
This work proposes a novel unsupervised entity alignment approach called UNEA, which achieves a new state-of-the-art for the unsupervised EA task, and can even outperform many existing supervised EA baselines.
- 2
- Optimizing traffic efficiency via a reinforcement learning approach based on time allocation2
A novel method is proposed that can naturally provide a countdown function by adopting a new action design that achieves control in the manner of time allocation, and the model figures out the duration of each phase at the beginning of every signal cycle, which is more practical for real-world traffic applications.
- 2
- General Rotation Invariance Learning for Point Clouds via Weight-Feature Alignment2
Weight-Feature Alignment (WFA) is proposed to construct a local Invariant Reference Frame (IRF) via aligning the features with the principal axes of the network weights to bridge the gap between rotation invariance learning and normal methods.
- 2
- 2
- 2
- Lyapunov‐Type Inequalities for Some Quasilinear Dynamic System Involving the (p1, p2, …, pm)‐Laplacian on Time Scales2
Several new Lyapunov-type inequalities for some quasilinear dynamic system involving the -Laplacian on an arbitrary time scale are established, which generalize and improve some related existing results including the continuous and discrete cases.
- 2
- Regularized Locality Preserving Projections with Two-Dimensional Discretized Laplacian Smoothing2
A Regularized LPP model using a Laplacian penalty to constrain the coefficients to be spatially smooth is introduced and by preserving the local geometrical structure of the image space, the model can obtain a linear subspace which is optimal for image representation in the sense of local isometry.
- 2
- M 3 CS: Multi-Target Masked Point Modeling With Learnable Codebook and Siamese Decoders1
M3CS introduces two decoders to reconstruct masked representations and the masked points simultaneously, and proposes an online codebook projecting continuous tokens into discrete ones before reconstructing masked points to compel the decoder to take effect through the combinations of tokens.
- Modeling Noncommutative Composition of Relations for Knowledge Graph Embedding1
A new KGE method called QuatRotatScalE (QRSE) is proposed, which utilizes rotation and scaling transformations of quaternions to design the relation embedding, and embeds the relations and entities into a quaternion vector space under the difference norm KGE framework.
- 1
- Denoising Multi-modal Sequential Recommenders with Contrastive Learning1
A weakly-supervised framework based on contrastive learning for denoising multi-modal recommenders (dubbed Demure) is proposed, which circumvents the requirement of users' explicit feedback and identifies the noises by analyzing the modalities of all interacted items from a given user.
- Towards In-distribution Compatibility in Out-of-distribution Detection1
This work identifies three causes of the in-distribution incompatibility: contradictory gradient, false likelihood, and distribution shift and proposes a new out-of-dist distribution detection method by adapting both the top-design of deep models and the loss function.
- 1
- Big data and machine learning techniques1
The two-volume set LNCS 9242 + 9243 constitutes the proceedings of the 5th International Conference on Intelligence Science and Big Data Engineering, IScIDE 2015, held in Suzhou, China, in June 2015.
- Event Recovery by Faster Truncated Nuclear Norm Minimization1
This paper forms an event as a two dimensional data matrix, which will be called the event matrix, and converts the original problem to matrix completion problem, and adopts a recently proposed approach called Truncated Nuclear Norm Minimization TNNM to recover theevent matrix.
- Orthogonal Projection Analysis1
This paper gives a continuous objective function based on gradient fields and discusses how to discretize it by using tangent space and shows the geometric meaning of the approach, which is requiring the gradient field as orthogonal as possible to the tangent spaces.
- 1
- 1
- Experimental Evaluation of Behavior Perception and Prediction Methods in Automated Port Logistics Operations–
This study shows that Probabilistic 3D significantly outperforms traditional methods in both detection and tracking accuracy for unmanned port logistics scenarios and makes autonomous logistics systems more reliable and robust in complex environments.
- Exercise Improved P2Y12-Regulated Microglial Dynamics During Stroke Via Endocannabinoid Signaling–
Endocannabinoid signaling may as the critical regulator of the PE-induced P2Y12-mediated effect whereby PE increased the endocannabinoids system to regulate purinergic signaling, further inducing microglial processes at microglia-neuron contacts to monitor and protect neuronal functioning.
- –
- –
- –
- –
- –
- –
- Stable Learning via Causality-based Feature Rectification.–
A novel Causality-based Feature Rectification (CFR) method to address the model misspecification problem under agnostic distribution shift by using a weight matrix to rectify features and can not only improve the stability of linear models, but also deep-learning based models.
- Partial Feature Decorrelation for Non-I.I.D Image classification–
A novel Partial Feature Decorrelation Learning (PFDL) Algorithm is presented, which jointly optimizes a feature decomposition network and the target image classification model, and can improve the backbone model's accuracy in non-i.i.d image classification datasets.
- Stable Learning via Self-supervised Invariant Risk Minimization–
This work proposes a novel Self-supervised Invariant Risk Minimization method based on the fact that the real causality connections between features are consistent no matter how the environment changes, which outperforms previous state-of-the-art stable learning methods.
- Editorial of the Special Issue on Manifold Learning.–
This special issue is to attract articles that address the frontier problems in the scientific principles of manifolds learning, and report empirical studies and applications of manifold learning algorithms, including but not limited to pattern recognition, computer vision, web mining, image processing and so on.
- –
- –
- A Geometric Perspective on Dimensionality Reduction–
Questions How can low dimensional structure in high dimensional data be detected and if so, what is its dimensionality?
- –
- –
- –
- –
- An Augmented Lagrangian Algorithm for Optimization with General Constraints–
The paper presents an augmented lagrangian algorithm for optimization with general constraints by applying to the function φ(x)=max｛g(x),-λr｝, which can be applied directly in the case of inequality constraints.
- –
- –
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 2026-10-14. 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.