More about HKUST
Computablizing Human Knowledge for Machines Through Domain-Specific Representation
PhD Qualifying Examination Title: "Computablizing Human Knowledge for Machines Through Domain-Specific Representation" by Mr. Yuzhe SHI Abstract: Domain practitioners working in critical sectors, including scientific discovery, engineering innovation, and social governance, are addressing frontier challenges within their respective domains by leveraging Domain-Specific Representations (DSRs) that render domain knowledge computable for Artificial Intelligence (AI) solutions. However, as domain practitioners craft DSRs tailored exclusively to their particular areas of focus, the diverse domain contexts and the methodologies of solving problems with that knowledge become intertwined. This results in fragmented, case-specific solutions with limited established practices to guide further research and development from a holistic AI application ecosystem viewpoint. This survey presents the first systematic analysis of DSR-AI approaches, disentangling domain context and methodology through a two-dimensional framework: (i) the functions of human knowledge, categorized according to Bayesian theory as the cognitive foundation, namely, knowledge with generative, discriminative, and descriptive functions; and (ii) the mechanisms by which human knowledge is utilized in AI, encompassing retrieval, modularization, composition, decomposition, planning, optimization, and explanation. Through this analytical framework, we illuminate the landscape, identify key challenges, and highlight research opportunities of knowledge utilization across domains, while elucidating shared methodologies for designing, deploying, and communicating DSRs to democratize AI for a broader spectrum of domains. Date: Tuesday, 26 August 2025 Time: 10:00am - 12:00noon Venue: Room 3494 Lifts 25/26 Committee Members: Prof. Huamin Qu (Supervisor) Dr. Shuai Wang (Chairperson) Dr. Jiasi Shen