2020-01-012024-05-13https://scholars.lib.ntu.edu.tw/handle/123456789/652514摘要：本中心開發智慧學習與人本資料分析技術用於資料的分析及運用，研究知覺計算與互動實境之知覺理解與反饋機制，提供更自然且直覺的互動模式，研發高速高效的可適性運算平台，有效地蒐集及擷取資料，並支援資料分析與虛擬實境所需之高效能運算，分成三個子計畫。子計畫一智慧學習技術與人本資料分析，發展從資料導向為基礎的機器學習技術，並融入保留使用者隱私的概念，一方面朝向人本智慧學習技術發展，另一方面期許能實現具社會價值與道德觀念之智能環境。子計畫二知覺計算與互動實境，旨在發展知覺資料分析與互動技術，讓機器理解知覺資料，並協助使用者有效利用，進一步提供更好的人機互動模式，處理語意鴻溝、巨量資料及隱私保護等挑戰，進而反饋知覺資料，藉由擴增與互動實境技術與真實感知資料融合，提供更好的互動體驗。子計畫三高速高效的可適性運算平台探究物聯網大數據應用情境中可組合系統平台上數據分析的架構設計、系統軟體、效能優化、軟硬整合、高速網路等關鍵議題，發展應用加值與軟硬整合技術。<br> Abstract: This center aims to develop advanced machine learning and human-centric data analytics techniques, design methods for perception data understanding and feedback, and deploy high-performance adaptive platforms for collecting, storing, and computing data. There are three sub-projects. The topic of the first sub-project is on machine learning techniques and human-centric data analytics. It starts with developing non-convex optimization algorithms, solving retrieval, and compressive sensing problems. Furthermore, the project will observe and derive knowledge graphs from such complex data to exploit the underlying semantic information and realize few/zero-shot learning tasks in supervised, semi-supervised, or even the challenging unsupervised settings. It will also emphasize the integration of privacy-preserving and moral/social values into the developed technologies, which would benefit users in real-world scenarios and result in intelligent environments. The second sub-project, perception computation and interactive reality, aims to develop analysis methods for perception data and interactive techniques for manipulating perception. The machines will better understand the context surrounding users, help users utilize multimedia data, and provide better human-machine interaction experience. The proposed methods have to address the challenges with such data, a significant semantic gap, a large volume of data, and privacy protection requirements. Users can effectively and securely analyze, search, and even generate data. Also, machines can better comprehend the surrounding around users, better understand users’ needs, and provide a better interactive experience. The third sub-project, a high-performance adaptive computation platform, develops the core technology of composable systems for data collecting, transmission, and computation. The research roadmaps cover data deduction on edge devices and cloud devices, performance optimization, heterogeneous computing, dynamic resource allocation, and optical communication among composable systems. They cover different aspects of data science and systems and focus on various stages of the data life cycle, acquisition, storage, analysis, utilization, and interaction. The center will contribute to academia and industry related to data science, human-computer interaction, artificial intelligence, and systems.數據科學人本資料分析計算知覺互動實境可適性運算平台Data ScienceHuman-centric data analyticsPerception ComputationInteractive realityHigh-performance adaptive platforms