William W. Cohen. Using distribution-free learning theory to analyze solutionpath caching mechanisms. Computational Intelligence, 8(2):336--375, 1992.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....in Minton s experiments and to informally explain several additional experimental results originally reported in [10] The experiments, and their analysis, provide a degree of confirmation for the structural theory. 7 Related Work A number of attempts have been made to analyze EBL s performance [6, 21, 28, 30, 44, 50]. Most of these analyses are very different from the analysis in this paper, and are discussed 7 RELATED WORK 36 0.00 500.00 1000.00 1500.00 2000.00 2500.00 Total CPU Time (Blocksworld) 0 20 40 60 80 100 120 140 160 Time Bound Prodigy Prodigy EBL Prodigy SUCCESS 0.00 500.00 1000.00 1500.00 ....

....and that, like inductive learning, EBL can be analyzed using Valiant s PAC framework [59] Although Tadepalli s model is different from my own, the spirit of the two approaches is similar: we both seek to identify classes of problem spaces in which EBL results in tractable problem solving. Cohen [6] describes a solution path caching mechanism that operates in polynomial time and provably improves performance (see also [5] Both Cohen and Tadepalli make useful connections between Valiant s framework and the learning of search control knowledge. Minton s thesis contains a thorough analysis of ....

William W. Cohen. Using distribution-free learning theory to analyze chunking. In Proceedings of the Eighth Biennial Conference of the Canadian Society for Computational Studies of Intelligence. Morgan Kaufmann, 1990.

....tree, where each node corresponds to a literal, and one descending arc indicates that this literal should be positive, and other, that is should be negative. 2] Several previous systems have implicitly dealt with this theme, of gaining efficiency at the expense of accuracy or categoricity; see [ Cohen, 1990 ] Subramanian and Genesereth, 1987 ] Levesque, 1986; Selman and Kautz, 1988; Etherington et al. 1989; Borgida and Etherington, 1989 ] Selman and Kautz, 1991 ] None, however, have explicitly quantified the tradeoffs. Improving accuracy (ignoring efficiency) One standard issue with ....

W. Cohen. Using distribution-free learning theory to analyze chunking. In CSCSI-90, 1990.

....produce a more efficient performance element. Most of the formal models, however, either deal with learning problems that are far harder than the problems actually attempted by real learning systems (e.g. GL89, Gre91] or model only relatively narrow classes of learning algorithms (e.g. NT88, Coh90] By contrast, our model is very general and directly relevant to many systems. There are also a great number of existing LfE systems, and considerable experimental work on the utility problem. Our research is not simply a retrospective analysis of these systems; it also augments that ....

W. Cohen. Using distribution-free learning theory to analyze chunking. In Proceedings of CSCSI-90, 1990.

....large branching factor. The peg solitaire domain cannot be handled efficiently by Micro Hillary since it does not use a small set of fixed operators. MacLearn solves it by using parameterized operators. Unlike some speedup learners that provide us with either statistical or theoretical guarantees (Cohen, 1992; Gratch DeJong, 1992; Greiner Likuski, 1989; Subramanian Hunter, 1992; Tadepalli Natarajan, 1996) Micro Hillary has a heuristic nature and does not provide us with any guarantee. Indeed, while it performs very well in some domains, it fails in other domains such as the N Hanoi. To handle ....

Cohen, W. W. (1992). Using distribution-free learning theory to analyze solution path caching mechanisms. Computational Intelligence, 8 (2), 336--375.

.... special cases of representation changes, most notably generating abstraction hierarchies [Korf, 1987; Knoblock, 1991b; Knoblock et al. 1991; Bacchus and Yang, 1992; Giunchiglia and Walsh, 1992] replacing operators with macros [Korf, 1985; Korf, 1987; Mooney, 1988] and learning control rules [Cohen, 1992]. A generalized model of representation changes was suggested by Korf, who formalized the concept of problem reformulation based on the notions of isomorphism and homomorphism of search spaces [Korf, 1980] Korf s model, however, does not address a method for evaluating the efficiency of a ....

William W. Cohen. Using distribution-free learning theory to analyze solution-path caching mechanisms. Computational Intelligence, 8(2):336--375, 1992.

....produce a more efficient performance element. Most of the formal models, however, either deal with learning problems that are far harder than the problems actually attempted by real learning systems (e.g. GL89, Gre91] or model only relatively narrow classes of learning algorithms (e.g. NT88, Coh90] By contrast, our model is very general and directly relevant to many systems. There are also a great number of existing LfE systems, and considerable experimental work on the utility problem. Our research is not simply a retrospective analysis of these systems; it also augments that ....

W. Cohen. Using distribution-free learning theory to analyze chunking. In Proceeding of CSCSI-90, 1990.

....require the learner to output an approximately correct problem solver with a high probability. Just as in PAC learning, we require the learner to be successful on any stationary problem distribution unknown to the learner. There have been some other attempts to formalize speedup learning (e.g. Cohen, 1992, Greiner Likuski, 1989, Subramanian Hunter, 1992) However, most of these formalizations of speedup learning use a measure of problem solving performance such as the number of nodes expanded in solving a problem (Cohen, 1992) or the number of unifications done in answering a query (Greiner ....

.... have been some other attempts to formalize speedup learning (e.g. Cohen, 1992, Greiner Likuski, 1989, Subramanian Hunter, 1992) However, most of these formalizations of speedup learning use a measure of problem solving performance such as the number of nodes expanded in solving a problem (Cohen, 1992) or the number of unifications done in answering a query (Greiner Likuski, 1989) We believe that these measures are too fine grained to be useful as a foundation for a robust theory of speedup learning comparable to the analysis of concept learning in the PAC learning framework. Following the ....

[Article contains additional citation context not shown here]

Cohen, W. (1992). Using distribution-free learning theory to analyze solution path caching mechanisms. Computational Intelligence, 8 (2), 336--375.

....loss, and also quantify the effects of abstraction on convergence. To make this paper more accessible to the general reader, proofs are either sketched or omitted in the body of the paper. Full proofs to all the theorems are included as an appendix. This paper is an extended version of [ Cohen, 1990 ] most of the technical material has previously appeared in [ Cohen, 1989 ] 2 Evaluating SLL Algorithms 2.1 Motivation of the Evaluation Criterion We first address the problem of how to evaluate the success of SLL algorithms in general, and in particular solution path caching algorithms. ....

William W. Cohen. Using distribution-free learning theory to analyze chunking. In Proceedings of the Eighth Biennial Conference of the Canadian Society for Computational Studies of Intelligence. Morgan Kaufmann, 1990.

No context found.

William W. Cohen. Using distribution-free learning theory to analyze solutionpath caching mechanisms. Computational Intelligence, 8(2):336--375, 1992.

No context found.

William W. Cohen. Using distribution-free learning theory to analyze solution-path caching mechanisms. Computational Intelligence, 8(2):336--375, 1992.

No context found.

W. W. Cohen. Using distribution-free learning theory to analyze solution path caching mechanisms. Computational Intelligence, 8(2):336--375, 1992.

No context found.

William W. Cohen. Using distribution-free learning theory to analyze solution-path caching mechanisms. Computational Intelligence, 8(2):336--375, 1992.

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