Empirical Methods for Artiicial Intelligence

Empirical Methods in AI gives researchers and practitioners a good introduction to exploratory data analysis (EDA), hypothesis testing, computer-intensive methods, and statistical techniques based on linear regression. The book motivates readers to use statistical techniques, which are explained through a large number of detailed examples. The basic ideas and the formulas one should use are explained, but the technical details were left to appendices and to other books. The practitioner will nd cookbook-style procedures that can be easily understood and utilized, but mathematically inclined readers will not nd any proofs of correctness nor a concise formal speci cation of the assumptions required to justify correctness of the procedures. Cohen (1991) surveyed the Eighth National Conference on Arti cial Intelligence (AAAI-90) and concluded that the methodologies used are incomplete with respect to the goals of designing and analyzing AI systems. Tichy, Lukowicz, Prechelt & Heinz (1995) showed in a very large study of over 400 articles that research papers in Computer Science are rarely validated with experimental results. More recently, and perhaps more appropriate for readers of this journal, Prechelt (1996) showed that the situation is not better in the neural network literature. Out of 190 articles published in well-known journals dedicated to neural networks, 29% did not employ even a single realistic or real learning problem. Only 8% of the articles presented results for more than one problem using real world data. While Prechelt (1996) only looked at whether comparisons were done, Cohen went a step further and described how to design good experiments. He wrote that \books like this one encourage well-designed experiments, which, if one isn't careful, can be utterly vacuous. This is a danger, not an inevitability. Knowing the danger, we can avoid it" (p. 103). As the title of the book implies, examples from Arti cial Intelligence are used throughout the book. The common urn and coinip examples from introductory Probability and Statistics textbooks were replaced by AI planners, expert systems, message understanding systems, and natural language problems. We see