国家科学基金会（National Science Foundation）研究生奖学金为本研究提供了部分赞助。本文中所表达的纯属作者的观点和结论，不应被明示或暗示地解释为代表美国国家科学基金会的官方政策。

　　参考文献

　　1. Berliner, H. and Ebeling, C., The SUPREM architecture: A new intelligent paradigm, Artificial Intelligence 28 (1986) 3-8.

　　2. Berliner, H., On the construction of evaluation functions for large domains, in: Proceedings IJCAI-79, Tokyo, Japan (1979) 53-55.

　　3. Condon, J.H. and Thompson, K., Belle chess hardware, in: M.R.B. Clark (Ed.), Advances in Computer Chess 3 (Pergamon Press, Oxford, 1982).

　　4. Duda, R. and Hart, P. Pattern Classification and Scene Analysis (Wiley, New York, 1973).

　　5. Griffith, A.K., A comparison and evaluation of three machine learning procedures as applied to the game of checkers, Artificial Intelligence 5 (1974) 137-148.

　　6. Hewlett, C., Report on a hardware computing system dedicated to the game of Othello, Othello Q. 8 (2) (1986) 7-8.

　　7. Lee, K. and Mahajan, S., BILL: A table-based knowledge-intensive Othello program, Carnegie-Melon University, Pittsburgh, PA (1986).

　　8. Mitchell, D., Using features to evaluate positions in experts' and novices' Othello games, Master Thesis, Northwestern University, Evanston, IL (1984).

　　9. Newell, A., Simon, H. and Shaw, C., Chess playing programs and the problem of complexity,

　　in: E.A. Feigenbaum and J. Feldman (Eds.), Computers and Thought (McGraw-Hill, New

　　York, 1963).

　　10. Pearl, J., Heuristics: Intelligent Search Strategies for Computer Problem Solving (Addison- Wesley, Reading, MA, 1984).

　　11. Rosenbloom, P.S., A world-championship-level Othello program, Artificial Intelligence 19 (1982) 279-320.

　　12. Samuel, A.L., Some studies in machine learning using the game of checkers, IBM J. 3 (1959) 210-229.

　　13. Samuel, A.L., Some studies in machine learning using the game of checkers, II, IBM J. 11 (1967) 601-617.

　　14. Slate, D.J. and Atkin, L.R., CHESS 4.6---The Northwestern University Chess Program, in:

　　Chess Skills in Man and Machine (Springer, Berlin, 1977) 101-107.

　　[1] 自动寻找好的特征是个非常棘手的问题，但是这不再本论文的讨论范围内。

　　[2] 每场游戏通常刚好进行60半步。唯一的例外是当双方都无合规棋步可走时。

　　[3] 由于随机初始化和终局搜索，算法未训练N < 24 和N > 49的位置。我们可以通过将N = 24的位置的参数复制到N < 24的位置上，将N = 49的位置的参数复制到N > 49的位置上。

　　[4] 我们将在第5.2.1小节证明这个假设是正确的。