Samuel's Checkers-Playing System
A Samuel's Checkers-Playing System is a Machine Learning-based checkers-playing program developed by Arthur Lee Samuel in ~1959.
- Example(s):
- …
- Counter-Example(s):
- See: Machine Learning and Game Playing, Reinforment Learning, Role Learning Task, Adaptive Non-Numeric Processing, Game Tree, Alpha-Beta Pruning, TD-Gammon.
References
2019
- (Wikipedia, 2019) ⇒ https://en.wikipedia.org/wiki/Arthur_Samuel Retrieved:2019-11-7.
- Arthur Lee Samuel (December 5, 1901 – July 29, 1990)[1] was an American pioneer in the field of computer gaming and artificial intelligence.[2] He coined the term “machine learning” in 1959. The Samuel Checkers-playing Program was among the world's first successful self-learning programs, and as such a very early demonstration of the fundamental concept of artificial intelligence (AI).[3] He was also a senior member in the TeX community who devoted much time giving personal attention to the needs of users and wrote an early TeX manual in 1983.[4]
- ↑ . A. Weiss (1992). “Arthur Lee Samuel (1901-90)". IEEE Annals of the History of Computing. 14 (3): 55–69. doi:10.1109/85.150082
- ↑ Samuel, Arthur L. (1959). “Some Studies in Machine Learning Using the Game of Checkers". IBM Journal of Research and Development. 44: 206–226. CiteSeerX 10.1.1.368.2254. doi:10.1147/rd.441.0206
- ↑ Gio Wiederhold; John McCarthy; Ed Feigenbaum (1990). "Memorial Resolution: Arthur L. Samuel" (PDF) . Stanford University Historical Society. Archived from the original (PDF) on 26 May 2011. Retrieved April 29, 2011.
- ↑ Donald Knuth (1990). "Arthur Lee Samuel, 1901-1990" (PDF). TUGboat. pp. 497–498. Retrieved April 29, 2011.
2017
- (Sammut & Webb, 2017) ⇒ Claude Sammut, and Geoffrey I. Webb. (2017). "Samuel's Checkers Player" In: (Sammut & Webb, 2017). DOI:10.1007/978-1-4899-7687-1_740
- QUOTE: Samuel’s Checkers Player is the first machine learning system that received public recognition. It pioneered many important ideas in game playing and machine learning. The two main papers describing his research (Samuel 1959, 1967) became landmark papers in Artificial Intelligence. In one game, the resulting program was able to beat one of America’s best players of the time.
(...) Samuel’s checkers player featured a wide variety of learning techniques. First, his checkers player remembered positions that it frequently encountered during play. This simple form of rote learning allowed it to save time, and to search deeper in subsequent games whenever a stored position was encountered on the board or in some line of calculation. Next, it featured the first successful application of what is now known as Reinforcement Learning for tuning the weights of its evaluation function. The program trained itself by playing against a stable copy of itself....
- QUOTE: Samuel’s Checkers Player is the first machine learning system that received public recognition. It pioneered many important ideas in game playing and machine learning. The two main papers describing his research (Samuel 1959, 1967) became landmark papers in Artificial Intelligence. In one game, the resulting program was able to beat one of America’s best players of the time.
2005
- (Sutton & Barton, 2005) ⇒ Richard S. Sutton and Andrew G. Barto (2005). "11.2 Samuel's Checkers Player" In: "Reinforcement Learning: An Introduction (Online Version)". Mark Lee 2005.
- QUOTE: Rote learning and other aspects of Samuel's work strongly suggest the essential idea of temporal-difference learning --that the value of a state should equal the value of likely following states. Samuel came closest to this idea in his second learning method, his “learning by generalization” procedure for modifying the parameters of the value function. Samuel's method was the same in concept as that used much later by Tesauro in TD-Gammon. He played his program many games against another version of itself and performed a backup operation after each move. The idea of Samuel's backup is suggested by the diagram in Figure 11.3. Each open circle represents a position where the program moves next, an non-move position, and each solid circle represents a position where the opponent moves next. A backup was made to the value of each on-move position after a move by each side, resulting in a second non-move position. The backup was toward the minimax value of a search launched from the second on-move position. Thus, the overall effect was that of a backup consisting of one full move of real events and then a search over possible events, as suggested by Figure 11.3. Samuel's actual algorithm was significantly more complex than this for computational reasons, but this was the basic idea.
Figure 11.3: The backup diagram for Samuel's checkers player.
- QUOTE: Rote learning and other aspects of Samuel's work strongly suggest the essential idea of temporal-difference learning --that the value of a state should equal the value of likely following states. Samuel came closest to this idea in his second learning method, his “learning by generalization” procedure for modifying the parameters of the value function. Samuel's method was the same in concept as that used much later by Tesauro in TD-Gammon. He played his program many games against another version of itself and performed a backup operation after each move. The idea of Samuel's backup is suggested by the diagram in Figure 11.3. Each open circle represents a position where the program moves next, an non-move position, and each solid circle represents a position where the opponent moves next. A backup was made to the value of each on-move position after a move by each side, resulting in a second non-move position. The backup was toward the minimax value of a search launched from the second on-move position. Thus, the overall effect was that of a backup consisting of one full move of real events and then a search over possible events, as suggested by Figure 11.3. Samuel's actual algorithm was significantly more complex than this for computational reasons, but this was the basic idea.
1996
- (Schaeffer & Lake, 1996) ⇒ Jonathan Schaeffer, and Robert Lake (1996). "Solving the game of checkers". Games of no chance, 29, 119-133.
- QUOTE: In the late 1950’s and early 1960’s, Arthur Samuel did pioneering work in artificial intelligence using the game of checkers as his experimental test bed [Samuel 1959; Samuel 1967]. Thirty years later, his work is still remembered, both for the significance of his research contributions, and for the legacy of his checkers-playing program. In 1962, Robert Nealy, blind checkers champion of Stamford, Connecticut, lost a single game to Dr. Samuel’s program. For the fledgling field of artificial intelligence, this event was viewed as a milestone and its significance was misrepresented in the media. Reporting of this event resulted in the game of checkers being labeled as “solved”: computers were better than all humans. Thus, a 1965 article by Richard Bellin in the Proceedings of the National Academy of Sciences declared that “ ... it seems safe to predict that, within ten years, checkers will be a completely decidable game” (vol. 53, p. 246), while Richard Restak, in the influential book The Brain: The Last Frontier (1979), stated that “ ... an improved model of Samuel’s checker-playing computer is virtually unbeatable, even defeating checkers champions foolhardy enough to ‘challenge’ it to a game” (p. 336).
1967
- (Samuel, 1967) ⇒ Arthur Lee Samuel (1967). "Some Studies In Machine Learning Using The Game Of Checkers II - Recent Progress". IBM Journal of Research and Development 11(6), 601-617.DOI: 10.1147/rd.116.0601
- QUOTE: Two machine learning procedures were described in some detail: (1) a rote learning procedure in which a record was kept of the board situation encountered in actual play together with information as to the results of the machine analyses of the situation; this record could be referenced at terminating board situations of each newly initiated tree search and thus, in effect, allow the machine to look ahead further than time would otherwise permit and, (2) a generalization learning procedure in which the program continuously re-evaluated the coefficients for the linear polynomial used to evaluate the board positions at the terminating board situations of a look-ahead tree search.
1959
- (Samuel, 1969) ⇒ Arthur Lee Samuel (1959). "Some Studies In Machine Learning Using The Game Of Checkers". IBM Journal of Research and Development 3(3):211–229. DOI: 10.1147/rd.33.0210