A5049810016- Advanced Graph Neural Networks
- Recommender Systems and Techniques
- Natural Language Processing Techniques
- Biomedical Text Mining and Ontologies
- Topic Modeling
- Complex Network Analysis Techniques
- Text and Document Classification Technologies
- Domain Adaptation and Few-Shot Learning
- Bioinformatics and Genomic Networks
- Data-Driven Disease Surveillance
- Semantic Web and Ontologies
- Machine Learning in Bioinformatics
- Epigenetics and DNA Methylation
- Imbalanced Data Classification Techniques
- Image and Video Quality Assessment
- Image Enhancement Techniques
- Knowledge Management and Sharing
- Multimodal Machine Learning Applications
- Intelligent Tutoring Systems and Adaptive Learning
- Color Science and Applications
- Music and Audio Processing
- Artificial Intelligence in Healthcare
- Automated Road and Building Extraction
- Functional Brain Connectivity Studies
- Mental Health Research Topics
Peking University
2021-2025
University of California, Los Angeles
2024
Fudan University
2010
The recently developed unsupervised graph representation learning approaches apply contrastive into graph-structured data and achieve promising performance. However, these methods mainly focus on augmentation for positive samples, while the negative mining strategies are less explored, leading to sub-optimal To tackle this issue, we propose a Graph Adversarial Contrastive Learning (GraphACL) scheme that learns bank of samples effective self-supervised whole-graph learning. Our GraphACL...
This article studies self-supervised graph representation learning, which is critical to various tasks, such as protein property prediction. Existing methods typically aggregate representations of each individual node representations, but fail comprehensively explore local substructures (i.e., motifs and subgraphs), also play important roles in many mining tasks. In this article, we propose a learning framework named cluster-enhanced Contrast (CLEAR) that models the structural semantics from...
In recent years, deep learning on graphs has achieved remarkable success in various domains. However, the reliance annotated graph data remains a significant bottleneck due to its prohibitive cost and time-intensive nature. To address this challenge, self-supervised (SSL) gained increasing attention made progress. SSL enables machine models produce informative representations from unlabeled data, reducing expensive labeled data. While witnessed widespread adoption, one critical component,...
Self-supervised graph representation learning has recently shown considerable promise in a range of fields, including bioinformatics and social networks. A large number contrastive approaches have promising performance for on graphs, which train models by maximizing agreement between original graphs their augmented views (i.e., positive views). Unfortunately, these methods usually involve pre-defined augmentation strategies based the knowledge human experts. Moreover, may fail to generate...
Graph neural networks (GNNs) have emerged as powerful tools for graph classification tasks. However, contemporary methods are predominantly studied in fully supervised scenarios, while there could be label ambiguity and noise real-world applications. In this work, we explore the weakly problem of partial learning on graphs, where each sample is assigned a collection candidate labels. A novel method called <underline xmlns:mml="http://www.w3.org/1998/Math/MathML"...
Collaborative Filtering (CF) is a pivotal research area in recommender systems that capitalizes on collaborative similarities between users and items to provide personalized recommendations. With the remarkable achievements of node embedding-based Graph Neural Networks (GNNs), we explore upper bounds expressiveness inherent methodologies, tackle challenges by reframing CF task as graph-signal processing problem. To this end, propose PolyCF, flexible graph signal filter leverages polynomial...
Semi-supervised node classification is a crucial challenge in relational data mining and has attracted increasing interest research on graph neural networks (GNNs). However, previous approaches merely utilize labeled nodes to supervise the overall optimization, but fail sufficiently explore information of their underlying label distribution. Even worse, they often overlook robustness models, which may cause instability network outputs random perturbations. To address aforementioned...
Source-free domain adaptation is a crucial machine learning topic, as it contains numerous applications in the real world, particularly with respect to data privacy. Existing approaches predominantly focus on Euclidean data, such images and videos, while exploration of non-Euclidean graph remains scarce. Recent neural network (GNN) can suffer from serious performance decline due shift label scarcity source-free scenarios. In this study, we propose novel method named Graph Diffusion-based...
In this paper, we study semi-supervised graph classification, which aims at accurately predicting the categories of graphs in scenarios with limited labeled and abundant unlabeled graphs. Despite promising capability neural networks (GNNs), they typically require a large number costly graphs, while wealth fail to be effectively utilized. Moreover, GNNs are inherently encoding local neighborhood information using message-passing mechanisms, thus lacking ability model higher-order dependencies...
Graph representation learning aims to effectively encode high-dimensional sparse graph-structured data into low-dimensional dense vectors, which is a fundamental task that has been widely studied in range of fields, including machine and mining. Classic graph embedding methods follow the basic idea vectors interconnected nodes can still maintain relatively close distance, thereby preserving structural information between graph. However, this sub-optimal due to: (i) traditional have limited...
Precisely defining the terminology is first step in scientific communication. Developing neural text generation models for definition can circumvent labor-intensity curation, further accelerating discovery. Unfortunately, lack of large-scale dataset hinders process toward generation. In this paper, we present a Graphine covering 2,010,648 pairs, spanning 227 biomedical subdisciplines. Terminologies each subdiscipline form directed acyclic graph, opening up new avenues developing graph-aware...
This paper studies the problem of modeling interacting dynamical systems, which is critical for understanding physical dynamics and biological processes. Recent research predominantly uses geometric graphs to represent these interactions, are then captured by powerful graph neural networks (GNNs). However, predicting in challenging scenarios such as out-of-distribution shift complicated underlying rules remains unsolved. In this paper, we propose a new approach named Graph ODE with...
Collaborative Filtering (CF) is a pivotal research area in recommender systems that capitalizes on collaborative similarities between users and items to provide personalized recommendations. With the remarkable achievements of node embedding-based Graph Neural Networks (GNNs), we explore upper bounds expressiveness inherent methodologies tackle challenges by reframing CF task as graph signal processing problem. To this end, propose PolyCF, flexible filter leverages polynomial filters process...
This paper studies the problem of class-imbalanced graph classification, which aims at effectively classifying categories graphs in scenarios with imbalanced class distribution. Despite tremendous success neural networks (GNNs), their modeling ability for graph-structured data is inadequate, typically leads to predictions biased towards majority classes. Besides, existing learning methods visions may overlook rich semantic substructures classes and excessively emphasize from minority To...
Recent years have witnessed the explosive growth of interaction behaviors in multimedia information systems, where multi-behavior recommender systems received increasing attention by leveraging data from various auxiliary such as tip and collect. Among recommendation methods, non-sampling methods shown superiority over negative sampling methods. However, two observations are usually ignored existing state-of-the-art based on binary regression: (1) users different preference strengths for...
The problem of matting is always solved by finding the alpha value for each pixel in image. Many recent methods combine color sampling and affinity definition different steps, leading to large computational cost. In proposed method, when a P <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i</sub> calculated, regarded as foreground help calculate its adjacent pixels' values, resulted faster solution. This spreading way traversal also ensures...
Precisely defining the terminology is first step in scientific communication. Developing neural text generation models for definition can circumvent labor-intensity curation, further accelerating discovery. Unfortunately, lack of large-scale dataset hinders process toward generation. In this paper, we present a Graphine covering 2,010,648 pairs, spanning 227 biomedical subdisciplines. Terminologies each subdiscipline form directed acyclic graph, opening up new avenues developing graph-aware...