Armand Prieditis and Jack Mostow. PROLEARN: towards a prolog interpreter that learns. In Proceedings of AAAI-87, AAAI, 1987.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....several concept recognition rules is simply to first try one, and then the other; this is equivalent to creating a new rule whose antecedent is the disjunction of the antecedents of the simple rules. This disjunctive rule can be simplified by combining common terms and or using partial evaluation [7, 11, 14, 16]; however, it is only in rare cases that the logical disjunction can be eliminated. This is a problem because disjunction acts like a backtracking point for the pattern matcher, increasing the cost of matching the antecedent of the rule. In Section 4, we explain how our technique allows rules to ....

....shape of the proof tree, rather than the specific values used in the proof tree. Typically this limitation shows up as an inability By standard EBL techniques I mean published domain independent explanation based learning algorithms, such as EGGS [15] PROLOG EBG [8] MRS EBG [6] and PROLEARN [16]. to generalize the number of entities involved in a proof or the number of times some action is performed; following [18] we refer to this phenomenon as generalizing number. As an illustration of generalizing number, consider a learning problem presented in [12] Here the training example is ....

Armand Prieditis and Jack Mostow. PROLEARN: towards a prolog interpreter that learns. In Proceedings of AAAI-87, AAAI, 1987.

....structure must terminate with an expression that satis es the operationality criterion. 2. determine a set of sucient conditions under which the explanation holds, stated in terms that satisfy the operationality criterion. In Logic Programming, EBG has been introduced by the works reported in [PM87], Hi87] KM87] and since then this kind of formalization has also been used to describe most of the previous domain independent characterizations of EBG. In the following, we will take [KM87] as base reference w.r.t. EBG. In order to relate this technique to the legal context, we will borrow ....

Prieditis A.E., Mostow J. PROLEARN: Towards a Prolog Interpreter that Learns. AAAI 87, 494-498.

....history to decide which state to backtrack to in case a search path proved fruitless. Explanation based learning (EBL) Mitchell 1997 ] takes the outcome of a training process (such as a rule, a proof or a decision tree) and transforms it to a more compact (and often more general) form. Prieditis and Mostow 1987 ] proposed an adaptive Prolog interpreter called PROLEARN which reduces the time of executing Prolog queries by using EBL to form generalizations of past proofs that are cached away and used in future problem solving episodes. Soar [ Laird et al. 1986 ] is a general problem solving architecture ....

A. E. Prieditis and J. Mostow. Prolearn: Towards a Prolog interpreter that learns. In Proceedings of the Sixth National Conference on Artificial Intelligence, Seattle, WA, 1987. Morgan Kaufmann.

....#ltering learned rules based on experimentation with those rules #Minton, 1988; Greiner Jurisica, 1992; Gratch Dejong, 1992; Markovitch Scott, 1993#. Heuristic approaches to generating learned rules have also been proposed that provide improved e#ciency over straightforward EBL such as #Prieditis Mostow, 1987; Minton, 1988; Shell Carbonell, 1991; Shavlik, 1990; Etzioni, 1990#. However, none of these approaches can guarantee that the cost of using the learned rules will always be bounded by the cost of the problem solving episode from which they are learned. That is, the cost of a learned rule can be ....

....the expensiveness of learned rules. One class of approaches to the expensive rule problem has focused on directly reducing the cost of learned rules. Some approaches have restructured and simpli#ed the learned rules to semantically equivalent ones in order to reduce the match cost of the rules #Prieditis Mostow, 1987; Minton, 1988#. Other approaches have analyzed the problem solving structure that is the basis of the learning to either avoid particular structures #such as recursion# in the learned rules #Etzioni, 1990#, or to preserve such structures in the learned rules #Shell Carbonell, 1991; Shavlik, ....

Prieditis, A. E. & Mostow, J. #1987#. PROLEARN: Towards a Prolog interpreter that learns. In Proceedings of the Sixth National ConferenceonArti#cial Intelligence, pages 494#498.

....history to decide which state to backtrack to in case a search path proved fruitless. ffl Explanation based learning (EBL) Mitchell 1997 ] takes the outcome of a training process (such as a rule, a proof or a decision tree) and transforms it to a more compact (and often more general) form. Prieditis and Mostow 1987 ] proposed an adaptive Prolog interpreter called PROLEARN which reduces the time of executing Prolog queries by referring to similar queries executed in the past. PROLEARN uses EBL to form generalizations of past proofs that are cached away 34 and used in future problem solving episodes. Soar [ ....

A. E. Prieditis and J. Mostow. Prolearn: Towards a prolog interpreter that learns. In Proceedings of the Sixth National Conference on Artificial Intelligence, Seattle, WA, 1987. Morgan Kaufmann.

....is capable of automatically transforming some intractable algorithms into ones that run in polynomial time. 1 Introduction Explanation based learning (EBL) research in logic programming has generally focussed on learning macros (compiled rules) Mitchell et al. 1986; DeJong and Mooney, 1986; Prieditis and Mostow, 1987 ] while EBL work in planning and production systems has tended to focus on learning search control rules [ Minton, 1988; Laird et al. 1986 ] Recently, Cohen [ Cohen, 1990 ] has argued the advantages of learning search control rules for the clause selection problem in logic programming. ....

A. Prieditis and J. Mostow. Prolearn: Towards a prolog interpreter that learns. In Proceedings of the Sixth National Conference on Artificial Intelligence, Seattle, WA, Jul 1987.

.... considered elsewhere (Ledeniov Markovitch, 1998a) Several researchers applied machine learning techniques for accelerating logic inference (Cohen, 1990; Dejong Mooney, 1986; Langley, 1985; Markovitch Scott, 1993; Minton, 1988; Mitchell, Keller, Kedar Cabelli, 1986; Mooney Zelle, 1993; Prieditis Mostow, 1987). Some of these works used explanation based learning or generalized caching tech76 The Divide and Conquer Subgoal Ordering Algorithm yes no no no yes yes yes no Average: 1.0 Test: bound(arg2) Average: 0.98 Test: bound(arg2) Average: 0.3 Test: female(arg1) Average: 0.0 Average: 0.0001 Average: ....

Prieditis, A. E., & Mostow, J. (1987). PROLEARN: Towards a prolog interpreter that learns. In Forbus, Kenneth; Shrobe, H. (Ed.), Proceedings of the 6th National Conference on Artificial Intelligence, pp. 494--498, Seattle, WA. Morgan Kaufmann.

....shown that it is possible to learn over one million rules while still allowing their efficient use (Doorenbos, Tambe, Newell 1992; Doorenbos 1993) In this article we focus on the expensive chunk problem. Previous work on the expensive chunk problem has investigated how to produce cheaper rules (Prieditis Mostow 1987; Minton 1988; Shell Carbonell 1991; Shavlik 1990; Etzioni 1990) and how to filter out expensive rules (Minton 1988; Greiner Jurisica 1992; Gratch Dejong 1992; Markovitch Scott 1993) However, none of these approaches can generally guarantee that the cost of using the learned rules will ....

Prieditis, A. E., and Mostow, J. 1987. PROLEARN: Towards a Prolog interpreter that learns. In Proceedings of the Sixth National Conference on Artificial Intelligence, 494--498.

....expensive. Recent work on the average growth effect has shown that it is possible to learn over one million rules while still allowing their efficient use [14, 15] This research focuses on the expensive chunk problem in EBL. There have been approaches which are useful for producing cheaper rules [16, 3, 12, 17, 11, 10] or filtering out expensive rules [3, 18, 19, 20] However, these approaches cannot generally guarantee that the cost of using the learned rules will always be bounded by the cost of the planning episode from which they are learned. That is, the cost of a learned rule can be greater than the cost ....

A. E. Prieditis and J. Mostow. Prolearn: Towards a prolog interpreter that learns. In Proceedings of the Sixth National Conference on Artificial Intelligence, pages 494--498, 1987.

....as a fruitful domain for EBL techniques virtually since their inception, since the notion of explanation may be usefully equated with the structure of a proof. Techniques for generalizing these structures and compiling macro rules are well known(Mitchell et al. 1986; DeJong and Mooney, 1986; Prieditis and Mostow, 1987). On the SBL side, there has been an explosion of recent research in inductive logic programming (ILP) which addresses the induction of Horn clause concept definitions (Muggleton, 1992; Quinlan, 1990) Third, expanding the search control framework into this new domain opens up a whole new range of ....

Prieditis, A. and Mostow, J. (1987). Prolearn: Towards a prolog interpreter that learns. In Proceedings of the Sixth National Conference on Artificial Intelligence. Seattle, WA.

....work on the average growth effect has shown that it is possible to learn over one million rules while still allowing their efficient use [10, 11] In this article we focus on the expensive chunk problem. Previous work on the expensive chunk problem has investigated how to produce cheaper rules [12, 5, 8, 13, 7] and how to filter out expensive rules [5, 14, 15, 16] However, none of these approaches can generally guarantee that the cost of using the learned rules will always be bounded by the cost of the problem solving episode from which they are learned. That is, the cost of a learned rule can be ....

A. E. Prieditis and J. Mostow. PROLEARN: Towards a Prolog interpreter that learns. In Proceedings of the Sixth National Conference on Artificial Intelligence, pages 494--498, 1987.

....and thus system can produce low quality solutions. 5 Related Work There has been a lot of work on the expensive chunk problem in EBL. Some approaches have restructured the learned rules to semantically equivalent ones to reduce the match cost of the rules. Partial evaluation in PROLEARN [25] simplifies the learned rules by exploiting domain constraints. COMPRESSOR [20] in Prodigy simplifies rules or combines multiple rules to find less expensive descriptions, by employing domain knowledge, partial evaluation, reordering, and logical equivalences. Although these restructuring ....

A. E. Prieditis and J. Mostow. Prolearn: Towards a prolog interpreter that learns. In Proceedings of the Sixth National Conference on Artificial Intelligence, pages 494--498, 1987.

....Many learning systems must confront the problem of run time after learning being greater than run time before learning. This utility problem has been a particular focus of research in explanation based learning (EBL) There have been approaches which are useful for producing cheaper rules [1, 2, 3, 4, 5, 6] or filtering out expensive rules [2, 7, 8, 9] However, these approaches cannot generally guarantee that the cost of using the learned rules will always be bounded by the cost of the problem solving from which they are learned, given the same situation. One way of finding a solution which can ....

A. E. Prieditis and J. Mostow. Prolearn: Towards a prolog interpreter that learns. In Proceedings of the Sixth National Conference on Artificial Intelligence, pages 494--498, 1987.

....framework that cleanly integrates these techniques can be achieved by considering the problem of control rule learning within the context of logic programming. 4. 2 Controlling Search in Logic Programs Although most EBL research in logic programming has generally focussed on learning macros, [25, 10, 36], Recent work has shown the utility of learning explicit search control rules within a logic programming framework [4, 55, 56] The execution of a logic program can be viewed as a problem solving process with a search strategy based on resolution theorem proving. A program executes by finding a ....

A. Prieditis and J. Mostow. Prolearn: Towards a Prolog interpreter that learns. In Proceedings of the Sixth National Conference on Artificial Intelligence, Seattle, WA, Jul 1987.

....system, and the information necessary to replicate the experiments in this paper, is available by sending mail to prodigy cs.cmu.edu. 1 INTRODUCTION 1 1 Introduction Explanation Based Learning (EBL) 11, 36] has emerged as a standard technique for acquiring search control knowledge (e.g. [27, 33, 37, 35, 47, 52]) 1 Recently, Prieditis [46] van Harmelen and Bundy [59] pointed to the similarity between Partial Evaluation (PE) a well known program optimization technique [25, 58] and EBL, suggesting that an EBL style analysis could be performed statically, without utilizing training examples. This ....

....WORK 30 part of the standard arsenal of PE systems. Finally, static anticipates goal clobbering and prerequisite violation whereas standard PE systems do not. In AI work, PE has been used to generate abstraction hierarchies for planning [7] and to reduce the match cost of rules learned via EBL [47, 29]. Concurrently with static, Letovsky [28] developed the prope system which generates macro clauses by partiallyevaluating pure PROLOG programs. Letovsky reports that prope failed to generate the appropriate search control knowledge for solving simple algebraic equations. prope is weaker than ....

[Article contains additional citation context not shown here]

A. E. Prieditis and J. Mostow. Prolearn: towards a Prolog interpreter that learns. In Proceedings of the National Conference on Artificial Intelligence. Morgan Kaufmann, 1987.

....can be slower after learning than before learning. Previous research on the utility problem has focused on detecting and discarding rules of low utility [ Minton, 1988; Markovitch and Scott, 1989 ] lowering the match cost of rules using partial evaluation or other simplification techniques [ Prieditis and Mostow, 1987; Minton, 1988; Tambe and Rosenbloom, 1989 ] and constraining the use of learned rules [ Mooney, 1989 ] This paper introduces two new techniques for improving the utility of learned rules. The first technique is to combine EBL with inductive learning techniques to learn a better set of ....

Armand Prieditis and Jack Mostow. PROLEARN: Towards a prolog interpreter that learns. In Proceedings of the Sixth National Conference on Artificial Intelligence, Seattle, Washington, 1987. Morgan Kaufmann.

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