有人能回答下面12个问题吗，我可以给600，
1. In which sense can computers "learn"? Describe the framework of concept learning.
2. Give some own example(s) of a learning problem (where we are in need of some results, but it is infeasible to obtain them in a direct way).
3. What is "generalization" beyond observed data, and has it something to do with the notion of a "more general hypothesis"?
4. Is it necessary to restrict the hypothesis space H, and how can (or should) it be done practically? Relate H to the notion of "inductive bias".
5. How is it possible at all for learning algorithms to make predictions about unseen data? To what extent can we trust their results?
6. Which computational difficulties can arise in concept learning tasks? How do various algorithms cope with these difficulties?
7. Does the ability to choose own training examples add power to a learner?
10. Characterize the concepts that are good to learn in the decision tree representation. Relate this matter to the inductive bias of decision tree learning.
11. What are the ideas behind using ensembles of hypotheses?
12. Which extensions of decision tree learning are in use, and how are they motivated?
13. Which types of concepts are good to learn by neural networks, and why?
联系方式，qq5459513
[返回] 计算机科学论坛 →
计算机理论与工程 → 『 人工智能 :: 机器学习|数据挖掘|进化计算 』 → 回答下面12题，给600
(订阅本版)
2008/9/6 11:45:00
新浪微博
回答下面12题，给600
