Contextual information plays an important role in action recognition. Local operations have difficulty to model the relation between two elements with a long-distance interval. However, directly modeling the contextua...
详细信息
The envisioned sixth-generation (6G) of wireless networks will involve an intelligent integration of communications and computing, thereby meeting the urgent demands of diverse applications. To realize the concept of ...
详细信息
Knowledge graph completion (KGC) aims to predict missing entities in knowledge graphs by learning effective representations of entities and their relations. Recent advances have explored multimodal KGC by incorporatin...
Knowledge graph completion (KGC) aims to predict missing entities in knowledge graphs by learning effective representations of entities and their relations. Recent advances have explored multimodal KGC by incorporating structural, textual, and visual information. However, two critical challenges remain unresolved: (1) modal heterogeneity, where significant differences in feature distributions across modalities hinder effective fusion; and (2) spatial heterogeneity, where embedding knowledge graphs in a single geometric space fails to capture their complex topological structures. To address these challenges, we propose ChoicE, a unified framework that leverages a mixture of experts (MoE) design for both encoding and decoding. In the encoder, the multimodal chooser preprocesses multiple modalities to derive embedding representations for each modality. These representations are then processed by distinct experts specialized for structural, textual, and visual features, facilitating effective fusion while preserving modality-specific information. In the decoder, the geometric chooser projects the unified multimodal embeddings into Euclidean, complex, or hyperbolic space, dynamically selecting the most appropriate space to model the inference patterns inherent to each query. Extensive experiments on multiple benchmark datasets demonstrate that ChoicE effectively overcomes these dilemmas and achieves state-of-the-art performance in multimodal KGC. The data and code are released at https://***/r/ChiocE-master/ .
Recently, significant research attention has been devoted to the study of reconfigurable intelligent surfaces (RISs), which are capable of reconfiguring the wireless propagation environment by exploiting the unique pr...
详细信息
In recent years, with the application and gradual popularization of UAV technology in many fields, the normalization of UAV aerial photography has become a common phenomenon. There are few studies on Multi-UAV regiona...
详细信息
Medical informatization takes a key role in medical and healthcare industries, which is a necessary way of improving service quality and treatment experience in a hospital. In this paper, we design and implement an in...
详细信息
Semantic Web (SW) has attracted the increasing attention of researchers, which facilitates people to link and handle various data. Ontology is the kernel technique of SW, and biomedical ontology is a state-of-art biom...
详细信息
ISBN:
(数字)9781728162157
ISBN:
(纸本)9781728162164
Semantic Web (SW) has attracted the increasing attention of researchers, which facilitates people to link and handle various data. Ontology is the kernel technique of SW, and biomedical ontology is a state-of-art biomedical knowledge modeling technique, which formally defines the biomedical concepts and their relationships. However, the same biomedical concepts in different biomedical ontologies could be defined in various contexts or with different terms, which yields the biomedical ontology heterogeneity problem. It is crucial to find mapping among heterogeneity concepts of different biomedical ontologies for bridging the semantic gaps, which is the so-called biomedical ontology matching. Biomedical ontology matching problem is an open challenge due to the rich semantic meaning and the flexible representation on a biomedical concept. To address this challenging problem, in this work, it is regarded as a binary classification problem, and a Long Short-Term Memory Networks (LSTM)-based ontology matching technique is proposed to solve it. Our proposal improves the quality of the alignment by introducing the char-embedding technique, which takes into account the semantic and context information of concepts. The comparing results with OAEI's participants show the effectiveness of our proposal.
Sentiment classification is a critical task in sentiment analysis and other text mining applications. As a sub-problem of sentiment classification, positive and unlab.led learning or positive-unlab.led learning (PU le...
详细信息
Nowadays, deep learning methods, especially the Graph Convolutional Network (GCN), have shown impressive performance in hyperspectral image (HSI) classification. However, the current GCN-based methods treat graph cons...
详细信息
Manifold learning has attracted more and more attention in machine learning for past decades. Unsupervised Large Graph Embedding (ULGE), which performs well on the large-scale data, has been proposed for manifold lear...
详细信息
暂无评论