Machine learning techniques for automated knowledge acquisition in intelligent knowledge-based systems.

摘要

The field of Artificial Intelligence addresses itself to ways in which intelligence can be imparted to machines and to what extent it can be done. Perhaps the most relevant aspect of intelligence in the context of machines is the ability to learn. Learning results in a qualitative and quantitative increase in the knowledge of a learner. This knowledge can then be used for future problem solving. In the realm of machines, questions about intelligence as determined by heredity and environment are not relevant, even though in the human context these questions have generated a lot of debate amongst sociologists and psychologists. Instead, Artificial Intelligence in general and Machine Learning (a subfield of Artificial Intelligence) in particular are concerned with architectural design and software systems which can facilitate the simulation of human intelligence and reasoning ability by machines.;This thesis investigates two important representative works from the field of Machine Learning to explore what lies ahead for the resolution of the problem of learning in machines. It is evident from the case study of research in Explanation-based Learning and Soar that there have been some very encouraging developments for the resolution of this problem. However, a lot more needs to be accomplished yet for the resolution of the problem of learning in machines.;Experimental, theoretical, and methodological progress made recently in the field of Machine Learning, has led to the ability to develop learning systems. These systems acquire expertise comparable to the best human expert knowledge in narrow task domains. It has led to new learning mechanisms based on using prior knowledge of the learner to reduce the difficulty of inductive inference in future problem solving. This progress has led to a significantly improved theoretical understanding of the computational limits of specific learning mechanisms. We now understand enough of the problem of Machine Learning to identify specific research directions and appropriate task domains to serve as driving forces for the next round of progress.;Extensive research effort is being expended to investigate and explore innovative ways of getting machines to learn from their experiences, to draw inferences based on inductive logic, to take remedial steps in case of flawed knowledge, and to initiate experimentation with the purpose of adding to or clarifying their domain knowledge. These aspects of learning are therefore appropriately the target of a major thrust of research in Artificial Intelligence because, it is primarily in these aspects that machine intelligence still lags well behind human intelligence. This research is a multi-disciplinary effort which draws upon the Cognitive Sciences, Artificial Intelligence, Psychology, Education, Philosophy, and other related disciplines.