|
Abstract : |
Machine learning techniques are perceived to have a great potential as means for the acquisition of knowledge; nevertheless, their use in complex engineering domains is still rare. Most machine learning techniques have been studied in the context of knowledge acquisition for well defined tasks, such as classification. Learning for these tasks can be handled by relatively simple algorithms. Complex domains present difficulties that can be approached by combining the strengths of several complementing learning techniques, and overcoming their weaknesses by providing alternative learning strategies. This study presents two perspectives, the macro and the micro, for viewing the issue of multistrategy learning. The macro perspective deals with the decomposition of an overall complex learning task into relatively well-defined learning tasks, and the micro perspective deals with designing multistrategy learning techniques for supporting the acquisition of knowledge for each task. The two perspectives are discussed in the context of, |
