Tadepalli, P. (1989). Lazy explanation-based learning: A solution to the intractable theory problem. Proceedings of the Eleventh International Joint Conferenceon Artificial Intelligence (pp. 694-700). Detroit, Michigan: Morgan Kaufmann.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....proven correct and complete control rules from a single (or few) problem solving examples and a correct underlying domain theory. They also require a complete domain theory to obtain the explanations, although there has been some work on learning with incomplete, or intractable theories, such as (Tadepalli, 1989). Alternatively, inductive ap proaches incrementally acquire correct knowledge by observing a large set of problem solving examples. These approaches strongly depend on the particular examples seen, but can also acquire simple and useful rules (Cohen, 1990, Leckie and Zukerman, 1991) This ....

....hypothesis. Some new techniques have been developed that use prior knowledge, but they are still mainly used for learning domain theories (Quinlan, 1990, Muggleton, 1992) instead of learning control knowledge. Similar work in lazy learning have been Lazy Explanation Based Learning, LEBL (Tadepalli, 1989), and Lazy Partial Evaluation, LPE (Clark and Holte, 1992) While LEBL refines the knowledge introducing exceptions, HAMLET modifies the control rules themselves by adding or removing their applicability conditions. Also, LEBL applies to games, while we use HAMLET for genera] task planning. ....

Prasad Tadepalli. Lazy explanation-based learning: A solution to the intractable theory problem. In Proceedings of the Eleventh International Joint Conference on Artificial Intelligence, pages 694-700, San Mateo, CA, 1989. Morgan Kaufmann.

.... organization essential and part of the overall learning process [4] 5 Related work Most speedup learning systems have been applied to problem solvers with the linearity assumption, such as the ones applied to Prolog or logic programming problem solvers [15, 21] special purpose problem solvers [12, 9, 18], or other general purpose linear problem solvers [5, 10, 11, 14] These problem solvers are known to be incomplete and and incapable of finding optimal solutions. If we remove the linearity assumption, we are dealing with nonlinear problem solvers. This kind of problem solvers are needed to ....

Prasad Tadepalli. Lazy explanation-based learning: A solution to the intractable theory problem. In Proceedings of the Eleventh International Joint Conference on Artificial Intelligence, pages 694--700, San Mateo, CA, 1989. Morgan Kaufmann.

....et al. 1986, Minton, 1988 ] usually involves a substantial effort to prove the correctness of the learned knowledge. In addition it requires a complete domain theory to obtain the explanations , although there have some work on learning with incomplete, or intractable theories, such as [ Tadepalli, 1989 ] Moreover, this approach does not work when dealing with a nonlinear planner [ Wang et al. 1993 ] The alternative approach, induction, usually requires a large set of examples and a long time to learn a correct description of the right control knowledge. Furthermore the method strongly ....

Prasad Tadepalli. Lazy explanation-based learning: A solution to the intractable theory problem. In Proceedings of the Eleventh International Joint Conference on Artificial Intelligence, pages 694--700, San Mateo, CA, 1989. Morgan Kaufmann.

....special purpose hardware. This model has been so successful that little else has been tried. The alternative AI approaches have not fared well due to the expense in applying the knowledge that had been supplied to the system. Those times in recent years that chess has been applied as a testbed [8, 27, 19, 21, 22, 30, 33, 26, 20] only a small sub domain of the game was used, so that fundamental efficiency issues that AI must grapple with have been largely unaddressed. However, we feel that there is a third approach that neither relies on search or the symbolic computation approach of knowledge oriented AI: what we shall ....

P. Tadepalli. Lazy explanation-based learning: A solution to the intractable theory problem. In Proceedings of the Eleventh International Joint Conference on Artificial Intelligence, Detroit, MI, 1989. Morgan Kaufmann. 13

.... #Wilensky, 1983#, limiting the maximum chain length #Mooney, 1990#, using heuristics to limit the branching factor of search #Hobbs et al. 1993#, using marker passing to propose candidate paths #Charniak, 1986; Norvig, 1989#, making simplifying assumptions about the explanations #Chien, 1989; Tadepalli, 1989#, and using plausibility estimates to guide the choice of which explanations to pursue #de Kleer Williams, 1989; Ng Mooney, 1990#. Nevertheless, the practical problem remains. The task of explaining anomalies accounting for reasoning #aws as well as accounting for surprising ....

Tadepalli, P. #1989#. Lazy explanation-based learning: a solution to the intractable theory problem. In Proceedings of the Eleventh International Joint Conference on Arti#cial Intelligence, pp. 694#700 Detroit, MI. IJCAI.

....learning, is on acquiring knowledge that improves the efficiency of planning. Several techniques have been used in this framework, including learning macro operators [Fikes et al. 1972, Korf, 1985, Cheng and Carbonell, 1.2. METHODOLOGY 3 1986, Segre, 1988] learning control rules [Minton, 1988, Tadepalli, 1989, Etzioni, 1990, Katukam and Kambhampati, 1994, Borrajo and Veloso, 1994a, Estlin and Mooney, 1996] learning by analogy [Veloso, 1994] chunking [Laird et al. 1986] and learning abstraction hierarchies [Knoblock, 1994, Christensen, 1990] The second category is on acquiring heuristics to guide ....

Prasad Tadepalli. Lazy explanation-based learning: A solution to the intractable theory problem. In Proceedings of the Eleventh International Joint Conference on Artificial Intelligence, pages 694--700, Detroit, MI, 1989.

....learning work by Flann (Flann and Dietterich, 1989) has occurred on only a very small sub domain of chess. The concepts capable of being learned by this system are graphs of two or three nodes in Morph. Such concepts are learned naturally by Morph s generalization mechanism. Tadepalli s work (Tadepalli, 1989) on hierarchical goal structures for chess is promising. Such high level strategic understanding may be necessary in the long run to bring Morph beyond an intermediate level (the goal of the current project) to an expert or master level. This brings out both a major weakness and a current topic of ....

P. Tadepalli. Lazy explanation-based learning: A solution to the intractable theory problem. In Proceedings of the Eleventh International Joint Conference on Artificial Intelligence, Detroit, MI, 1989. Morgan Kaufmann.

....a deductive approach and invest a substantial explanation effort to produce proven correct and complete control rules from a single (or few) problem solving examples and a correct and complete underlyingdomain theory. There has been work on learning with incomplete, or intractable theories, e.g. (Tadepalli 1989)) applied to simple problem solving scenarios. Alternatively to these deductive domain theory dependent algorithms, inductive learning approaches incrementally acquire correct knowledge by observing a large set of 1 HAMLET stands for Heuristics Acquisition Method byLearning from sEarch Trees ....

Tadepalli, P. 1989. Lazy explanation-based learning: A solution to the intractable theory problem. In Proceedings of the Eleventh International Joint Conference on Artificial Intelligence, 694--700. San Mateo, CA: Morgan Kaufmann.

....goal from a definite (i.e. Horn clause) domain theory. However, it is a fact that practical applications frequently require the enhanced expressiveness of negations in rule bodies. Specifically, this is the case for the domain of game playing, where traditional EBG has turned out to be inadequate [Ta89]. In this paper we present an approach which extends EBG to this more general setting; it is described in the form of a transformation system, and comprises Siqueira and Puget s method of Explanation Based Generalization of Failures [SiPu88] for definite programs. For the case that both domain ....

.... Part of them, as e.g. Minton s Constraint Based Generalization [Min84] deduce correct recognition rules for different games, but are limited to certain classes of positions (in this case to those for which a forcing strategy exists) Others (e.g. Flann and Dietterich [FlaDi86] Tadepalli [Ta89], and Yee et al. YSUB90] derive incorrect conditions which are subsequently refined by different methods. The intractable theory problem [Ta89] states that traditional EBG is not directly applicable, although a complete axiomatization can be provided. We conjecture that one of the underlying ....

[Article contains additional citation context not shown here]

Tadepalli, P., Lazy Explanation-Based Learning: A Solution to the Intractable Theory Problem, Proc. IJCAI-89, Detroit MI, (1989), 694-700

....learning, namely, Foil. The first use of approximations in learning control rules was probably MetaLEX [ Keller, 1987 ] However, it used a fairly simple method of simplifying learned rules by removing conditions. Most other recent investigations in learning approximations [ Ellman, 1988; Tadepalli, 1989; Chien, 1989 ] have not focussed on search control heuristics. Approximating control rules was investigated in [ Chase et al. 1989 ] however, their system, ULS, does not employ induction and is therefore limited to conservative approximations. Reported improvements in efficiency for ULS were ....

Tadepalli, P. (1989). Lazy explanation-based learning: A solution to the intractible theory problem.

....position [50] The symbolic learning work by Flann [8] has occurred on only a very small sub domain of chess. The concepts capable of being learned by this system are graphs of two or three nodes in Morph. Such concepts are learned naturally by Morph s generalization mechanism. Tadepalli s work [47] on hierarchical goal structures for chess is promising. We suspect that such high level strategic understanding may be necessary in the long run to bring Morph beyond an intermediate level (the goal of the current project) to an expert or master level. Minton [30] building on Pitrat s work [35] ....

P. Tadepalli. Lazy explanation-based learning: A solution to the intractable theory problem. In Proceedings of the Eleventh International Joint Conference on Artificial Intelligence, Detroit, MI, 1989. Morgan Kaufmann.

....an example is a member of target concept, and then confirms the conjecture with empirical data. Other systems have employed lazy explanation based learning (LEBL) which generates incomplete explanations and then incrementally refines any overly general knowledge using new planning examples (Tadepalli, 1989; Borrajo Veloso, 1994b) This proposal presents a novel multi strategy learning approach to control knowledge acquisition. Our learning system Scope also uses a combination of EBL and induction to learn control information. However, instead of generating control rules through EBL and then ....

Tadepalli, P. (1989). Lazy explanation-based learning: A solution to the intractable theory problem. In Proceedings of the Eleventh International Joint Conference on Artificial Intelligence Detroit, MI.

.... Third, because it will learn successfully in a domain using an intractable theory, it will constitute a solution to the intractable theory problem, a problem that is being pursued by researchers in explanation based learning [Mitchell, Keller Kedar Cabelli, 1986; Mostow Fawcett, 1987; Tadepalli, 1989]. Fourth, because features are abstractions of states, the system constitutes an automatic transformation based abstraction system. All other such systems require a human to control the application of the transformation. This research, using feedback from an inductive concept learner to control ....

Tadepalli, P. (1989). Lazy explanation-based learning: A solution to the intractable theory problem. Proceedings of the Eleventh International Joint Conference on Artificial Intelligence (pp. 694-700). Detroit, Michigan: Morgan Kaufmann.

.... will appear quite often in this paper, we want to stress that the general principles exemplified by this domain are valid for all complex, intractable domains (e.g. reasoning about physical systems [Weld and deKleer, 1990] medical domains [Bratko et al. 1989] music [Widmer, 1992b] chess [Tadepalli, 1989], Flann, 1989] etc. The next three chapters give a review of previous work in Machine Learning on the topics of approximation, learning with qualitative and plausible theories and abstraction. Very loosely these chapters correspond to the three different kinds of abstraction as seen in [Doyle, ....

....be judged quite accurately. For these reasons chess has been called the drosophila 2 of AI alluding to the important role this comparably simple animal played as the object of early research in genetics [McCarthy, 1990] 3 In particular chess has become a standard example [Tadepalli, 1986, Tadepalli, 1989, Flann, 1990] for an intractable domain. 2.1 Knowledge is intractable Many applications of rule based production systems assume that the underlying theory is complete and consistent. However, real world domains usually can t be formalized in a nice and neat way [Rajamoney and DeJong, 1987] It ....

[Article contains additional citation context not shown here]

Prasad Tadepalli. Lazy explanation-based learning: A solution to the intractable theory problem. In Proceedings of the 11th International Joint Conference on Artificial Intelligence, 1989.

....learn rules has been developed : Explanation Based Learning (EBL) Mitchell 1986] Dejong 1986] This learning method is particularly useful in the domain of games. Games have a strong domain theory. Many projects using EBL to learn tactical plans have been developed [Minton 1984] Puget 1987] [Tadepalli 1989] after the initial work of Jacques Pitrat on Chess [Pitrat 1976] This article explains a method to learn using fuzzy explanations. I have developed a systems which learns tactical plans in the game of Go. It learns using explanations on the problems is has solved [Cazenave 1996b] My system also ....

- P. Tadepalli. Lazy Explanation-Based Learning: A Solution to the Intractable Theory Problem. IJCAI 1989, pp. 694-700, 1989.

....rules derived by his program need not be sufficient for deriving the target concept (as it is usually the case in explanation based learning) In other words the learned patterns suggest good moves, but do not guarantee the successful outcome of the combination in slightly different situations. Tadepalli (1989) has also acknowledged this problem and proposed a different solution. The philosophy of Lazy Explanation Based Learning is to attack this problem by paying no attention to the possible refutations of a move. The program learns a set of over general, optimistic plans called o plans. All moves in ....

Tadepalli, P. (1989). Lazy explanation-based learning: A solution to the intractable theory problem. In Proceedings of the 11th International Joint Conference on AI, pp. 694--700. Morgan Kaufmann.

....by previous interesting experiences. In contrast, McCallum (1995) explained how storing case his8 tories of exploration traces can help to solve state aliasing problems. Analytic learning: Some research on explanation based learning has examined the utility of lazy approaches for deduction (e.g. Tadepalli (1989); Clark Holte (1992) More recently, Borrajo and Veloso (1997) have shown that a lazy learning approach can be used to efficiently acquire control knowledge in an incremental planning process. 2.1.4 Summary This section briefly described investigations of research related to CBR in a few ....

Tadepalli, P. (1989). Lazy explanation-based learning: A solution to the intractable theory problem.

....of recent progress in relational learning, namely, Foil. The first use of approximations in learning control rules was probably MetaLEX (Keller, 1987) which used a simple technique for removing conditions. Most other recent investigations have not focussed on learning control rules (Ellman, 1988; Tadepalli, 1989; Chien, 1989) or have not employed induction (Chase et al. 1989) Yoo and Fisher (Yoo and Fisher, 1991) combine induction and explanation to improve performance in a problem solving framework. They enhance the utility of EBL macros by clustering them in a Cobweb style classification tree ....

Tadepalli, P. (1989). Lazy explanation-based learning: A solution to the intractable theory problem.

....learning work by Flann [ Flann and Dietterich, 1989 ] has occurred on only a very small sub domain of chess. The concepts capable of being learned by this system are graphs of two or three nodes in Morph. Such concepts are learned naturally by Morph s generalization mechanism. Tadepalli s work [ Tadepalli, 1989 ] on hierarchical goal structures for chess is promising. We suspect that such high level strategic understanding may be necessary in the long run to bring Morph beyond an intermediate level (the goal of the current project) to an expert or master level. Minton [ Minton, 1984 ] building on ....

P. Tadepalli. Lazy explanation-based learning: A solution to the intractable theory problem. In Proceedings of the Eleventh International Joint Conference on Artificial Intelligence, Detroit, MI, 1989. Morgan Kaufmann. 18 6. Conclusions and Ongoing Directions

....only more recently, casebased reasoning (Kolodner, 1993; Aamodt Plaza, 1994) and machine learning (Aha et al. 1991) Second, almost all address supervised learning tasks. The exception is by Borrajo and Veloso, which is in the tradition of research on lazy analytic learning algorithms (e.g. Tadepalli, 1989; Clark Holte, 1992; Bostrom, 1992) Third, most of the articles focus on partially lazy learning algorithms (i.e. they compromise on some of the distinguishing characteristics of purely lazy algorithms noted above) While purely lazy algorithms are useful for some tasks, there are many ways in ....

Tadepalli, P. (1989). Lazy explanation-based learning: A solution to the intractable theory problem.

....of machine learning to planning systems has focused on planning efficiency, that is, on acquiring problem solving strategies that control search in order to make problem solving more efficient. This area of research has been termed speed up learning [ Mitchell et al. 1986; Minton et al. 1989; Tadepalli, 1989; Gratch et al. 1993; P erez and Etzioni, 1992 ] The research described in this paper looks instead at the application of machine learning to acquire strategies that lead a planner towards improving plan quality, an essential step in transforming planners from research tools into real world ....

....quality metric. Note that such control knowledge is orthogonal to planning efficiency control knowledge for early pruning of choices that are guaranteed to lead to failure paths. Such control knowledge has been the target of extensive research [ Mitchell et al. 1986; Minton et al. 1989; Tadepalli, 1989; Gratch et al. 1993; P erez and Etzioni, 1992 ] Figure 2 shows the architecture of quality which addresses the quality mapping problem. quality is given a domain theory D (operators and inference rules) and a domain dependent metric that evaluates the quality of the plans produced QD (P ) ....

Prasad Tadepalli. Lazy explanation-based learning: A solution to the intractable theory problem. In Proceedings of the Eleventh International Joint Conference on Artificial Intelligence, pages 694--700, Detroit, MI, 1989.

....precursor of such approaches to learning in games. Meanwhile, the knowledge representation of these programs make them learn overly general knowledge. This obliges them, either to partially verify their assertions, or to use their learned knowledge as heuristics. This also the case in the work of [Tadepalli 1989] on lazy EBL: his program learns overgeneral rules and refines them. Unlike these programs, my approach is to learn only certain rules. In order to achieve this goal, I explicitly represent uncertainty and separate clearly what is uncertain from what is certain. I do not want to verify the ....

Tadepalli P. (1989). Lazy Explanation-Based Learning: A Solution to the Intractable Theory Problem. IJCAI 1989, pp. 694-700, 1989.

.... (Wilensky, 1983) limiting the maximum chain length (Mooney, 1990) using heuristics to limit the branching factor of search (Hobbs et al. 1993) using marker passing to propose candidate paths (Charniak, 1986; Norvig, 1989) making simplifying assumptions about the explanations (Chien, 1989; Tadepalli, 1989), and using plausibility estimates to guide the choice of which explanations to pursue (de Kleer Williams, 1989; Ng Mooney, 1990) Nevertheless, the practical problem remains. The task of explaining anomalies accounting for reasoning flaws as well as accounting for surprising ....

Tadepalli, P. (1989). Lazy explanation-based learning: a solution to the intractable theory problem.

No context found.

Tadepalli, P. (1989). Lazy explanation-based learning: A solution to the intractable theory problem. Proceedings of the Eleventh International Joint Conferenceon Artificial Intelligence (pp. 694-700). Detroit, Michigan: Morgan Kaufmann.

No context found.

Prasad Tadepalli. Lazy explanation-based learning: A solution to the intractable theory problem. In Proceedings of the Eleventh International Joint Conference on Artificial Intelligence, pages 694--700, San Mateo, CA, 1989. Morgan Kaufmann.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC