A5091407514- Data Quality and Management
- Multimedia Communication and Technology
- Semantic Web and Ontologies
- Research Data Management Practices
- Scientific Computing and Data Management
- Advanced Graph Neural Networks
- Peer-to-Peer Network Technologies
- Power Systems and Technologies
- Privacy-Preserving Technologies in Data
- Mobile Agent-Based Network Management
- Advanced Database Systems and Queries
- Access Control and Trust
- Graph Theory and Algorithms
- Robotics and Automated Systems
- Biomedical Text Mining and Ontologies
- Service-Oriented Architecture and Web Services
- Logic, Reasoning, and Knowledge
- Reinforcement Learning in Robotics
- Time Series Analysis and Forecasting
- Image and Object Detection Techniques
- AI-based Problem Solving and Planning
- Wireless Sensor Networks and IoT
- Topic Modeling
- Neural Networks and Applications
- Digital Holography and Microscopy
Vrije Universiteit Amsterdam
2019-2024
Beijing University of Civil Engineering and Architecture
2023
Hangzhou Dianzi University
2021-2023
State Grid Corporation of China (China)
2022
Hubei University of Technology
2021
Chinese Academy of Sciences
2014-2021
University of Chinese Academy of Sciences
2014-2021
Shenyang Institute of Computing Technology (China)
2021
National University of Defense Technology
2019
Zhejiang University
2011-2019
Automatic emotion recognition is a challenging task which can make great impact on improving natural human computer interactions. In this paper, we present our effort for the Affect Subtask in Audio/Visual Emotion Challenge (AVEC) 2017, requires participants to perform continuous prediction three affective dimensions: Arousal, Valence and Likability based audiovisual signals. We highlight aspects of solutions: 1) explore fuse different hand-crafted deep learned features from all available...
Various benchmarks have been proposed to assess the performance of large language models (LLMs) in different coding scenarios. We refer them as code-related benchmarks. However, there are no systematic guidelines by which such a benchmark should be developed ensure its quality, reliability, and reproducibility. propose How2Bench, is comprised 55- 55-criteria checklist set govern development comprehensively. Using HOW2BENCH, we profiled 274 released within past decade found concerning issues....
This study reviewed the use of Large Language Models (LLMs) in healthcare, focusing on their training corpora, customization techniques, and evaluation metrics. A systematic search studies from 2021 to 2024 identified 61 articles. Four types corpora were used: clinical resources, literature, open-source datasets, web-crawled data. Common construction techniques included pre-training, prompt engineering, retrieval-augmented generation, with 44 combining multiple methods. Evaluation metrics...
FAIR Implementation Profiles (FIPs) are created and published to capture decisions made by communities on data curation management. Despite that many FIPs were in domains such as medicine environmental science, few available social sciences. This extended abstract reports recent advances creating using It consists of a summary Three use cases included demonstrate how can be used guide management for researchers, organisations, communities. Finally, we envision some future work FIP development.
While Deep Neural Networks (DNNs) achieve state-of-the-art performance in many fields, e.g., object recognition, they rely on deep networks with millions or even billions of parameters. Accelerating DNNs by reducing the parameters is crucial for real-time recognition. This paper presents an evolutionary approach to evolve efficient that can be run Low-Performance Computing Hardware (LPCH) recognition fastest possible speed and accuracy more than 95%. achieves goal means two design choices....
Large knowledge graphs capture information of a large number entities and their relations. Among the many relations they capture, class subsumption assertions are usually present expressed using \texttt{rdfs:subClassOf} construct. From our examination, publicly available contain potentially erroneous cyclic subclass relations, problem that can be exacerbated when different integrated as Linked Open Data. In this paper, we an automatic approach for resolving such cycles at scale automated...
Aiming at the assembly work of robot base and first axis gear reducer in industrial body production line, this paper proposed a visual recognition algorithm based on improved Hough transform circle detection algorithm. The could accomplish rapid positioning workpiece with high precision. Combined application heavy-load robot, whole system realized intelligent reducer. That is "manufacturing robot". Finally, feasibility reliability verified by experiment.
This paper studies artistic expression in human movement by exploring the performance art form salsa. The motions of a salsa are constructed as concatenations motion primitives, each which specifies dance pair over course eight musical beats. To analyze syntax expression, choreography performances is represented transition model that based on humanoid robot representations dancers. In order to assess quality performance, two distinct metrics explored. By integrating into proposed system, it...
Global registration is an important step in the three-dimensional reconstruction of multi-view laser point clouds for moving objects, but severe noise, density variation, and overlap ratio between present significant challenges to global registration. In this paper, a cloud method based on low-rank sparse decomposition proposed. Firstly, spatial distribution features were extracted by rasterization realize loop-closure detection, corresponding weight matrix was established according...
Due to problems in manufacturing process and forging technology, some of the workpieces production workshop have defects, which affects ornamental value utility workpieces. At present, detection workpiece defects is still mainly relying on manual detection, consumes human resources has a high error rate. To solve above problems, deep learning used improve reduce cost production. This article uses Faster R-CNN1 as basic architecture integrates Resnet502 backbone network. In addition,...
With the development of intelligent war, unmanned vehicle plays an more and important role in future war. As part ground system, combat vehicles have been a hot spot for researchers to improve autonomy. knowledge is foundation intelligence great significance improving This paper proposes build graph domain. The firstly uses ontology pattern layer graph. In order fully express uncertainty domain knowledge, probabilistically extended support representation uncertain knowledge. Then reasoning...
Action learning is an important aspect of self-improving. This paper explores a new approach for the two types actions, namely precondition-free actions and conditional actions. The corresponding algorithms are designed implemented using modern logic reasoners. Finally, simple system action agents to explore cooperative self-improving multi-agent systems.
Deep learning-based Autonomous Driving (AD) models often exhibit poor generalization due to data heterogeneity in an ever domain-shifting environment. While Federated Learning (FL) could improve the of AD model (known as FedAD system), conventional struggle with under-fitting amount accumulated training progressively increases. To address this issue, instead small models, employing Large Vision Models (LVMs) is a viable option for better learning representations from vast volume data....
Evolutionary Algorithms (EAs) have been widely used in several fields due to their ease of implementation and effectiveness dealing with complicated problems. Most algorithms attempt guide search direction by leveraging information from superior solutions within the population, often referred as "positive information." However, they frequently overlook value using data inferior solutions—termed "negative information"—to avoid less promising areas. In a scenario that exclusively considers...
Abstract One of the key critical technologies in digital revolution measurement technology is twin. The literature now publication indicates that advancement and use twin will raise bar for improvement measuring sector. current on creation reviewed first, followed by a list recognized definitions summary three main categories models easy reference. drawbacks conventional application process are enumerated here: direct challenging, multiple parameters at once sensors’ influence cannot be...
The mission of resilience Ukrainian cities calls for international collaboration with the scientific community to increase quality information by identifying and integrating from various news social media sources. Linked Data technology can be used unify, enrich, integrate data multiple In our work, we focus on datasets about damaging events in Ukraine due Russia's invasion between February 2022 end April 2023. We convert two selected enrich them additional geospatial information. Following...
Smartphone agents are increasingly important for helping users control devices efficiently, with (Multimodal) Large Language Model (MLLM)-based approaches emerging as key contenders. Fairly comparing these is essential but challenging, requiring a varied task scope, the integration of different implementations, and generalisable evaluation pipeline to assess their strengths weaknesses. In this paper, we present SPA-Bench, comprehensive SmartPhone Agent Benchmark designed evaluate...