ARTICLES ALPHA-BETA PRUNING: THEY'VE SEEN IT BEFORE Joseph S. Fulda Hofstra University Hempstead. Long Island, New York 11550 One of the m o s t apparently subtle t e c h n i q u e s taught in a typical AI survey course is minimaxing with a l p h a beta pruning. As Winston writes in his text: "It is not u n usual to get lost in this .... Even seasoned g a m e specialists still feel magic in the a l p h a - b e t a p r o c e d u r e Each i n dividual c o n c l u s i o n seems right, but s o m e h o w the global result is strange," One way of lessening the appar ent m y s t e r y of the t e c h n i q u e is to g r o u n d it in s o m e t h i n g v i r tually every c o m p u t e r science student has seen and m a s tered earlier: c u t o f f of evaluation in conditionals. After e n c o u n t e r i n g a true disjunct or a false conjunct in thre p rotasis of an IF s t a t e m e n t a r e a s o n a b l e c o m p i l e r will cease evaluation and return true and false respectively. It is easy to see by c o n s t r u c t i n g some quick truth tables that the MAX function and disjunction are the same. Likewise, the MIN function and conjunction are equivalent. To stop e v a l u a t i o n of a MAX function, what occurs in a l p h a - b e t a pruning, is thus no d i f f e r e n t f r o m w h a t happens to e v e r y d a y c o m p o u n d c o n d i t i o n s with a disjunction. Mutatis mutandis, the same thing can be said of the MIN f u n c t i o n and conjunction. There is one slight difference, however. Ordinary c o n d i t i o n a l s are based on standard, t w o - s t a t e logic, while the MAX and MIN f u n c t i o n s in m i n i m a x i n g can take on many values. We thus have to c o m p l e t e our c o m p a r i s o n by noting that for m u l t i - v a l u e d logics a true s t a t e m e n t is as true as its truest disjunct, while a false s t a t e m e n t is as false as its falsest conjunct. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . evaluation of atomic and m o l e c u l a r data (Cann and Nicholls, 1980), it was decided to i n c o r p o r a t e names of researchers, dates of data added as well as critical n o t e s on m e t h o d s and r e c o m m e n d a t i o n s on results. The reasons f o r this w e r e to be able to c o m p a r e and update e x p e r i m e n t a l data with time and with new e x p e r i m e n t a l results. But, new results are not a l w a y s better. Often. older values are o b t a i n e d t h r o u g h m o r e t h o r o u g h study then s o m e n e w e r results. There are several d i f f e r e n t types of errors that can exist in data or k n o w l e d g e bases: 1. Poor quality of i n f o r m a t i o n in basic e x p e r i m e n t a l data 2. Poor quality of i n f o r m a t i o n due to subjective i n t e r pretation 3. Erroneous i n f o r m a t i o n f r o m n o n - e x p e r t s 4. Time d e p e n d e n t data (related values and tools change) 5. Factual i n f o r m a t i o n due to m i s - i n t e r p r e t a t i o n or t y p o g r a p h i c a l errors 6. M i s - i n f o r m a t i o n - i n f o r m a t i o n d e s i g n e d to lead one astray for a specific p u r p o s e (often military o r political) 7. Malicious i n f o r m a t i o n e n t e r e d r a n d o m l y All of t hese c a t e g o r i e s must be faced in t r y i n g to maintain the k n o w l e d g e base of an e x p e r t s y s t e m in p r o d u c t i o n use. This p r o b l e m is in addition to the p r o b l e m of c o n s i s t ency of the data. For a small e x p e r t s y s t e m of less than 150 rules, the m a i n t e n a n c e may not present a severe problem. However, f o r a large system, revalidating and maintaining c o n s i s t e n c y should be very difficult. The maj or f u n c t i o n s p e r f o r m e d in m a i n t e n a n c e of k n o w l e d g e bases should be: 1. Changi ng w e i g h t s on rules due to r e f i n e m e n t o f knowledge. 2. Replacing k n o w l e d g e that is f o u n d to be incorrect, i n a p p r o p r i a t e or inaccurate. 3. Discarding of k n o w l e d g e 4. Updating of time i n d e p e n d e n t i n f o r m a t i o n 5. Adding of new c o n c l u s i o n s (new diseases, etc) 6. Adding new rules 7. Adding or replacing c o m m e n t s In a p r o d u c t i o n business e n v i r o n m e n t , this m a i n t e n a n c e would likely be done by multiple, changing p r o g r a m m e r s and "experts" of various quality. The mai nt enan c e should be done with a backtracking capability such that the c o m pany can r e - c r e a t e its expert s y s t e m to any state for business or legal requirements. The expert system should also be validated af t er any set of changes is applied Knowledge system m a i n t e n a n c e should be easier Cf the f o l l o w i n g i n f o r m a t i o n is kept in the system for each piece of knowledge: 1. Source of the i n f o r m a t i o n (individual, publication, etc) 2. Date that the knowledge was d e t e r m i n e d or published) 3. C o m m e n t s on the information, and possible i n t e r actions In the normal usage of the expert system, this m a i n tenance i n f o r m a t i o n is not necessary. However, in adding M a i n t e n a n c e of Expert Systems: Life-Cycle Validity John C. McCallum D e p a r t m e n t of I n f o r m a t i o n Systems and C o m p u t e r Science National University of S i n g a p o r e Singapore 0511 It is well k n o w n in the business c o m p u t i n g c o m munity that p r o g r a m and data base maintenance are the largest c o m p o n e n t of the s o f t w a r e system life cycle (Boehm, 1981). Little concern has been raised a b o u t the m a i n t e n a n c e of k n o w l e d g e bases of expert systems. The majority o f interest in m a i n t e n a n c e (which might be c l a s sified as being validity over the life cycle of the expert system (Mostow, 1985), and in the m a i n t e n a n c e of c o n s i s tency of a data base (Balzer et al, 1983). One c o m m e r c i a l selling strategy of e x p e r t s y s t e m s is that they do not change with time like human k n o w l e d g e . A person w h o doesn't change with time h o w e v e r , is not necessarily an intelligent person. An expert s y s t e m w i t h o u t m a i n t e n a n c e is "an old fogey," unable to learn new t h i n g s In a p r e l i m i nary study on h o w to i m p l e m e n t a data base for critical SIGART Newsletter, October 1985, N u m b e r 94 Page 26

/lp/association-for-computing-machinery/alpha-beta-pruning-they-ve-seen-it-before-MOaIE1fPtc