Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks96 from public data
- Multimodal Sentiment Detection Based on Multi-channel Graph Neural Networks167
This paper proposes Multi-channel Graph Neural Networks with Sentiment-awareness (MGNNS) for image-text sentiment detection and introduces multi-channel graph neural networks to learn multimodal representations based on the global characteristics of the dataset.
- COSTA124
COSTA, a novel Covariance-preServing feaTure space Augmentation framework for GCL is proposed, which generates augmented features by maintaining a "good sketch" of original features, and it is shown that the feature augmentation with COSTA achieves comparable/better results than graph augmentation based models.
- A Survey of Trustworthy Federated Learning: Issues, Solutions, and Challenges94
This work charts out a comprehensive road-map for Trustworthy Federated Learning (TFL) and provides an overview of existing efforts across four pivotal dimensions: Privacy and Security, Robustness, Fairness, and Explainability.
- A Survey of Trustworthy Federated Learning with Perspectives on Security, Robustness and Privacy90
This survey outlines the threats that pose vulnerabilities to trustworthy federated learning across different stages of development, including data processing, model training, and deployment, and discusses specific technical solutions for realizing each aspect of Trustworthy FL (TFL).
- 84
- #DebateNight: The Role and Influence of Socialbots on Twitter During the 1st 2016 U.S. Presidential Debate71
It is found that many highly influential Twitter users are in fact pro-Democrat and that most pro-Republican users are mid-influential and likely to be human (low botness), which allows for a nuanced analysis of the interplay between botness, partisanship and influence.
- Low-Rank Adaptation for Foundation Models: A Comprehensive Review67
This survey provides the first comprehensive review of LoRA techniques beyond large Language Models to general foundation models, including recent techniques foundations, emerging frontiers and applications of low-rank adaptation across multiple domains.
- 60
- Recent Advances of Multimodal Continual Learning: A Comprehensive Survey56
The first comprehensive survey on MMCL is presented, providing essential background knowledge and MMCL settings, as well as a structured taxonomy of MMCL methods, which categorize MMCL methods into four categories, i.e., regularization-based, architecture-based, replay-based, and prompt-based methods.
- Few-shot Joint Multimodal Aspect-Sentiment Analysis Based on Generative Multimodal Prompt45
This work proposes a novel Generative Multimodal Prompt (GMP) model for MABSA, which includes the Multinodal Encoder module and the N-Stream Decoders module, and introduces a subtask to predict the number of aspect terms in each instance to construct the multimodal prompt.
- Towards an Optimal Asymmetric Graph Structure for Robust Semi-supervised Node Classification45
A probabilistic model is proposed to infer the edge weights in this graph, which can be jointly trained with the subsequent node classification component and the effectiveness of the method and the rationality of the optimal graph is demonstrated.
- MM-InstructEval: Zero-Shot Evaluation of (Multimodal) Large Language Models on Multimodal Reasoning Tasks35
This paper conducts extensive zero-shot evaluations on 45 models (including 36 MLLMs) across 16 multimodal datasets, encompassing 6 distinct tasks using 10 different instructions and introduces multiple innovative metrics, including the 'Best Performance' metric to benchmark peak model capabilities.
- Graph-adaptive Rectified Linear Unit for Graph Neural Networks35
This work proposes Graph-adaptive Rectified Linear Unit (GReLU), which is a new parametric activation function incorporating the neighborhood information in a novel and efficient way and conducts comprehensive experiments to show that the plug-and-play GReLU method is efficient and effective given different GNN backbones and various downstream tasks.
- 34
- Contrastive Cross-scale Graph Knowledge Synergy33
To promote cross-scale information sharing and knowledge interactions among GP, a novel joint optimization formula that contains a pairwise contrastive loss between any two coarse-grained graph views is proposed that promotes knowledge sharing that yields informative representations, but also stabilizes the training process.
- Semi-supervised Multi-label Learning for Graph-structured Data33
This work incorporates the idea of label embedding into the proposed model to capture both network topology and higher-order multi-label correlations and demonstrates that the model significantly outperforms the state-of-the-art models for node-level tasks.
- 33
- Bipartite Graph Convolutional Hashing for Effective and Efficient Top-N Search in Hamming Space32
This work proposes an end-to-end Bipartite Graph Convolutional Hashing approach, namely BGCH, which consists of three novel and effective modules: adaptive graph convolutional hashing, latent feature dispersion, and Fourier serialized gradient estimation.
- oneDNN Graph Compiler: A Hybrid Approach for High-Performance Deep Learning Compilation30
OneDNN Graph Compiler is presented, a tensor compiler that employs a hybrid approach of using techniques from both compiler optimization and expert-tuned kernels for high-performance code generation of the deep neural network graph.
- AnnaAgent: Dynamic Evolution Agent System with Multi-Session Memory for Realistic Seeker Simulation29
AnnaAgent is proposed, an emotional and cognitive dynamic agent system equipped with tertiary memory that achieves more realistic seeker simulation in psychological counseling compared to existing baselines.
- Dual-view hypergraph neural networks for attributed graph learning28
A Dual-view HyperGraph Neural Network (DHGNN) model for attributed graph learning that unify the expression form of different information sources of nodes by hypergraph, and construct dual hypergraphs according to topology and attributes of nodes.
- Doc2hash: Learning Discrete Latent variables for Documents Retrieval25
This paper proposes a method, Doc2hash, that solves the gradient flow problem of the discrete stochastic layer by using continuous relaxation on priors, and trains the generative model in an end-to-end manner to generate hash codes.
- 23
- Geometric View of Soft Decorrelation in Self-Supervised Learning22
This work provides a new perspective on the geometric distance between positive definite matrices to investigate why the soft decorrelation cannot efficiently solve the dimensional collapse and constructs a family of loss functions utilizing the Bregman Matrix Divergence.
- WSFE: Wasserstein Sub-graph Feature Encoder for Effective User Segmentation in Collaborative Filtering21
This work proposes WSFE, a model-agnostic and training-free representation encoder, to be flexibly employed on the fly for effective user segmentation and demonstrates the superiority and generality of WSFE to fuel multiple downstream tasks with diverse underlying targets in recommendation.
- InterSentiment: combining deep neural models on interaction and sentiment for review rating prediction20
A novel, extensible review rating prediction model called InterSentiment is proposed by bridging the user-product interaction model and the sentiment model based on deep learning, and is a specific instance of the proposed Deep Learning based Collaborative Filtering framework.
- Can Watermarked LLMs be Identified by Users via Crafted Prompts?19
This work designs an identification algorithm called Water-Probe that detects watermarks through well-designed prompts to the LLM, and introduces the Water-Bag strategy, which significantly improves watermark imperceptibility by merging multiple watermark keys.
- Understanding and Mitigating Hyperbolic Dimensional Collapse in Graph Contrastive Learning16
A novel contrastive learning framework to learn high-quality graph embeddings in hyperbolic space is proposed and the alignment metric that effectively captures the hierarchical data-invariant information, as well as a substitute of the uniformity metric to prevent the so-called dimensional collapse is proposed.
- TOOL-ED: Enhancing Empathetic Response Generation with the Tool Calling Capability of LLM15
This paper proposes Emotional Knowledge Tool Calling (EKTC) framework, which encapsulates the commonsense knowledge bases as empathetic tools, enabling LLMs to integrate external knowledge flexibly through tool calling.
- 15
- Graph Component Contrastive Learning for Concept Relatedness Estimation15
This paper introduces a novel data augmentation approach to sample new concept pairs from the graph, and formalizes the CRE properties and introduces a graph structure named ConcreteGraph, which satisfies a set of intrinsic properties of CRE.
- An Effective Post-training Embedding Binarization Approach for Fast Online Top-K Passage Matching14
This paper proposes an effective post-training embedding binarization approach over ColBERT, achieving both architecture-level and embedding-level optimization for online inference, and demonstrates the efficaciousness of this approach to perform online query-passage matching acceleration.
- 12
- Towards Effective Top-N Hamming Search via Bipartite Graph Contrastive Hashing12
This work investigates the problem of hashing with Graph Convolutional Network for effective Top-N search and proposes Bipartite Graph Contrastive Hashing (BGCH+), a novel dual augmentation approach to both intermediate information and hash code outputs in the latent feature spaces that produces more expressive and robust hash codes within a dual self-supervised learning paradigm.
- Contrastive Learning with Generated Representations for Inductive Knowledge Graph Embedding12
The proposed VMCL proposes representation generation to capture the encoded and generated representations of entities, where the generated variations can augment the representation space with complementary features and design two CL objectives that work across entities and meta-KGs to simulate the transfer mode.
- Utilizing Computational Music Analysis and AI for Enhanced Music Composition: Exploring Pre- and Post-Analysis11
The synergistic roles of pre-analysis and post-analysis techniques in leveraging AI-driven tools to enhance the creative process and quality of musical compositions are explored.
- 11
- 11
- 10
- CSPO: Cross-Market Synergistic Stock Price Movement Forecasting with Pseudo-volatility Optimization10
This work introduces the framework of Cross-market Synergy with Pseudo-volatility Optimization (CSPO), a deep neural architecture to leverage external futures knowledge that enriches stock embeddings with cross-market insights and thus enhances the CSPO's predictive capability.
- 10
- 10
- High-precision three-dimensional imaging based on binocular meta-lens and optical clue fusion9
The innovation lies in the synergistic fusion of physics-derived absolute stereo depth measurements and machine learning-estimated relative depth through adaptive confidence mapping - the latter effectively addressing the inherent limitations of absolute depth estimation in scenarios with insufficient matching features.
- FedHLT: Efficient Federated Low-Rank Adaption with Hierarchical Language Tree for Multilingual Modeling8
This work introduces a communication-efficient framework for Multilingual Modeling that combines low-rank adaptation with a hierarchical language tree structure and mitigate parameter conflicts by organizing languages based on their familial ties rather than merging all LoRA parameters together.
- 8
- Interpretable Emotion Analysis Based on Knowledge Graph and OCC Model7
An analysis framework is constructed for interpreting casual association based on the emotional logic and the results demonstrate that this framework can make better inferences on emotions and good interpretability for human beings.
- 6
- 6
- 6
- 6
- MDAL: Multi-task Dual Attention LSTM Model for Semi-supervised Network Embedding6
A Multi-task Dual Attention LSTM model (dubbed as MDAL), which can capture structure, content, and label information of network and adjust representation vectors according to the concrete downstream task simultaneously is proposed.
- Learning distributed representations for community search using node embedding6
This paper proposes to exploit node embedding technique for community search task from a new perspective to learn a mapping from nodes to points in a lowdimensional vector space, which can be used as feature inputs for downstream data mining models.
- Rationality and Connectivity in Stochastic Learning for Networked Coordination Games5
This work considers a task allocation problem as a binary networked coordination game over an undirected regular graph, and shows that there is a relationship between the connectivity of the network, and the rationality of the agents.
- 5
- GvdsSQL: Heterogeneous Database Unified Access Technology for Wide-Area Environments4
This paper implements a unified data access mechanism for heterogeneous databases through metadata extraction, abstracts the unified access model, and accomplishes identification and forwarding of fundamental database operations, and implements a multilevel caching mechanism for query results in wide-area databases utilizing semantic analysis.
- DialogConv: A Lightweight Fully Convolutional Network for Multi-view Response Selection4
This paper proposes a novel lightweight fully convolutional architecture, called DialogConv, for response selection, exclusively built on top of convolution to extract matching features of context and response.
- 4
- 4
- Towards Non-Euclidean Foundation Models: Advancing AI Beyond Euclidean Frameworks3
This workshop focuses on the intersection of Non-Euclidean Foundation Models and GEometric Learning (NEGEL), exploring its potential benefits, including the potential benefits for advancing web-related technologies, challenges, and future directions.
- ConSurv: Multimodal Continual Learning for Survival Analysis3
This work proposes ConSurv, the first multimodal continual learning (MMCL) method for survival analysis, and introduces a new benchmark integrating four datasets, Multimodal Survival Analysis Incremental Learning (MSAIL), for comprehensive evaluation in the CL setting.
- On the role of network structure in learning to coordinate with bounded rationality3
It is shown that when using a network of irrational agents, promoting more homogeneous connectivity improves the accuracy of collective decision-making, as predicted by the phenomenon known as “Wisdom of Crowds.”
- Graph neural network for recommendation in complex and quaternion spaces3
This paper proposes Graph Collaborative Filtering for recommendation in Complex and Quaternion space (GCFC and GCFQ respectively), which uses complex embeddings or quaternion embeddings to initialize users and items and uses the Hermitian or Hamilton product with inner product to obtain the intricate antisymmetric relations between users and items.
- PersonaGAN: Personalized Response Generation via Generative Adversarial Networks3
A generative adversarial network model PersonaGAN is proposed that can generate more coherent and personalized responses while ensuring fluency, and a speaker-aware encoder is introduced in the generator to capture richer context information.
- PrivRec: User-Centric Differentially Private Collaborative Filtering Using LSH and KD3
PrivRec is a user-centric differential private collaborative filtering approach that provides privacy guarantees both intuitively and theoretically while preserving recommendation utility and the experimental results show that the approach provides promising utility with privacy guarantees compared to the commonly used approaches.
- Teaching Effectiveness and Student Experience in University Interactive Learning Platforms Based on Mobile Technology2
Empirical evidence is provided for the application and extension of TAM in the field of educational technology and offers theoretical and practical implications for university administrators in optimizing digital education strategies, for educators in designing effective interactive learning activities, and for platform developers in enhancing system design.
- Track and Tweak: Monitoring and Improving Group Fairness for Temporal Graph Neural Networks in Real Time2
This work is the first to investigate group fairness on temporal graphs and propose a novel real-time framework for monitoring and improving group fairness in Temporal Graph Neural Networks, named FTGNN-M, which offers theoretical guarantees for both fairness assessment and fairness promotion.
- 2
- Variational autoencoder densified graph attention for fusing synonymous entities: Model and protocol2
- 2
- 2
- 2
- 2
- 2
- Towards effective privacy preservation in federated learning with automatic gradient clipping and gradient transformation perturbation1
A novel privacy preservation scheme for federated learning that combines automatic gradient clipping and gradient transformation perturbation that achieves a good trade-off between data privacy and effectiveness under the same privacy budget is proposed.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- Context-aware Non-electronic Service Discovery and Composition1
This thesis provides an approach for using contexts to discover and compose non-electronic services, and presents a new context model which is to make it more suitable for service composition.
- FLASH: Fast Generative Retrieval via Autoregressive Semantic Hashing with Provably Distance Bounds–
This work proposes Flash (Fast Generative Retrieval via Autoregressive Semantic Hashing), a framework that integrates random orthogonal projection hashing (ROPH) with autoregressive hash code generation and introduces a dual-head autoregressive decoder that jointly predicts hash tokens and a quantization scalar.
- When Gradient Boosting Meets Adapter: Exploring Weak Learners for Parameter-Efficient Fine-tuning of LLMs–
This work proposes eXtreme Gradient Boosting LoRA (XGBLoRA), a novel framework grounded in gradient boosting theory that provides theoretical analysis establishing convergence guarantees and expressiveness bounds, which formally justify why weaker (lower-rank) adapters, when properly combined, can match or exceed the performance of stronger ones.
- Federated Domain Generalization for Time-Series Classification via Dynamics-to-Domain Generation–
This work proposes FedD2G, a dynamics-aware generative framework for federated domain generalization that introduces Sequence-Fused Transition sampling with a class-conditioned Markov transition model to assemble descriptors into temporally coherent sequences, enabling controlled synthesis that expands beyond observed source distributions without sharing raw signals.
- LC-ERD: Mining Latent Logic for Self-Evolving Reasoning via Consistency-Regulated Reward Decomposition–
Experiments show LC-ERD delivers a robust self-evolution path, uncovering trade-offs between logic consistency and accuracy while identifying high-value reasoning patterns missed by standard rewards.
- –
- –
- Stock Price Prediction using LSTM Model–
The author uses the Long-Short Term Memory (LSTM) network to make predictions by predicting the real stock prices of Google, Amazon, and Apple, and gives the visualization results of the training and testing stages.
- RoCar: A Relationship Network-based Evaluation Method for Large Language Models–
The RoCar method is proposed, which utilizes the defined basic schemas to randomly construct a task graph and generates natural language evaluation tasks based on the task graph to evaluate the reasoning and memory abilities of LLMs respectively.
- –
- Research on Unreasonable Parameter Identification Model of Overhead Line Based on k-means Clustering Algorithm–
The general characteristics and methods of power grid planning are summarized, the theory of data mining is introduced, and an unreasonable parameter identification model for overhead transmission lines based on-means clustering algorithm is established.
- –
- –
- –
- Research on E-commerce payment system–
This paper emphasizes on the research of payment system in E-commerce and a more scientific and safe payment is expected.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.