Theory of Learning Algorithms

这项研究是关于学习算法的理论研究，它使用了来自教师的例子和附加信息。它是由专家系统中的知识获取问题引起的。考虑的附加信息包括对教师的询问和教师设计的培训序列，即旨在指导和简化学习过程的类似或相关现象的例子。将“可能近似正确”的概念推广到仅从正例中学习，并研究了它与伪随机数生成器的关系。

This research deals with the theoretical investigation of learning algorithms that use examples and additional information from a teacher. It is motivated by the problem of knowledge acquisition in expert systems. The additional information considered includes queries to the teacher and training sequences devised by the teacher, that is, examples of similar or related phenomena intended to guide and simplify the learning process. The extension of the "probably approximately correct" concept to learning from positive-only examples, and the study of its relation to pseudo-random number generators are also proposed.