作者: Scott B. Huffman , Douglas J. Pearson , John E. Laird
DOI: 10.1007/978-1-4615-3172-2_6
关键词: Domain theory 、 Variety (cybernetics) 、 Focus (optics) 、 Descriptive knowledge 、 Computer science 、 Contrast (statistics) 、 Theoretical computer science 、 Domain (software engineering) 、 Task (project management) 、 Imperfect
摘要: Explanation-Based Learning (Mitchell et al., 1986; DeJong and Mooney, 1986) has shown promise as a powerful analytical learning technique. However, EBL is severely hampered by the requirement of complete correct domain theory for successful to occur. Clearly, in non-trivial domains, developing such nearly impossible task. Therefore, much research been devoted understanding how an imperfect can be corrected extended during system performance. In this paper, we present characterization problem, use it analyze past area. Past characterizations problem (e.g, Rajamoney DeJong, 1987)) have viewed types performance errors caused faulty primary. contrast, focus primarily on knowledge deficiencies theory, from these derive that result. Correcting search through space possible theories, with variety sources used guide search.