Craig A. Knoblock. Automatically Generating Abstractions for Problem Solving.PhD thesis, School of Computer Science, Carnegie Mellon University, 1991. Available as Technical Report CMU-CS-91-120.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... These methods in clude macro operator formation [Korf 7 19857 Iba 7 19897 Minton 7 19857 Cheng and Carbonell 7 19867 Shell and Carbonell 7 198917 explanationbased learning [Mitchell et al. 7 19867 DeJong and Mooney 7 19867 Minton et al. 7 1989a] 7 abstraction [Sacerdoti 7 19777 Korf 7 19877 Knoblock 7 1991] and within domain analogy [Carbonell 7 19837 Carbonell 7 19867 Veloso and Carbonell 7 19897 Hickman et al. 7 1990] The time has come to address machine learning in the large 7 in cluding both for inductive concept acquisition [Catlett 7 19917 Quinlan 7 1987] and analytic performance ....

.... various learning techniques: explanation based learning (EBL) Minton, 1988] acquisition of control knowledge through static analysis [Etzioni, 1990] learning by analogy [Veloso, 1991] learning by experimentation [Carbonell and Gil, 1990] learning abstraction hierarchies for effective planning [Knoblock, 1991], and semiautomated knowledge acquisition interfaces [Joseph, 1989] These techniques have been developed and tested in a variety of small and medium domains, and all exhibit varying degrees of improved performance. Addressing large scale problems, however, requires several types of analyses and ....

[Article contains additional citation context not shown here]

Knoblock, C. A. (1991). Automatically Generating Abstractions for Problem Solving. PhD thesis, School of Computer Science, Carnegie Mellon University, Pittsburgh, PA. Available as technical report CMU-CS-91-120.

.... Chien [3] talks about the need for tools in domain development, including tools for the development of operators, for planning systems to compare favourably in terms of software life cycle costs to other means of automation , Currie and Tate say it is hard work [5] Knoblock a black art [6] and McCluskey talks about the coding up and maintenance of these operators as difficult [7] Our experience also suggests that operator development is the most difficult part of writing a new domain. Planner operators have a dual aspect on the one hand they embody knowledge about the ....

C.A. Knoblock "Automatically Generating Abstractions for Problem Solving" Ph.D. Thesis, School of Computer Science, Carnigie Mellon University, Pittsburgh, PA (1991).

....of abstraction. The only states that can be abstracted are those states shared with the next more abstract level, so there is a search within a level of abstraction to find a state that can be abstracted. Also, the abstract search completely obviates the need for lower level search. Knoblock [Knoblock, 1991] uses a more traditional planning search space. He shows that abstraction can reduce an exponential planning search to linear with appropriate abstractions, but that result only holds under a set of strong assumptions. We also use a traditional planning search space, but in contrast to Knoblock, ....

....at a lower level of abstraction, the operator effects produce intermediate goals, which may thus overlap temporally or even be out of sequence. Because of this lack of strict ordering among intermediate goals, the search at the specific level cannot use the goals as milestones in the sense of [Knoblock, 1991]. Knoblock assumes that the search can achieve one intermediate goal completely independently from all previous and subsequent goals. This is a problem in general, even without our lack of strict temporal sequence (for instance, the Sussman anomaly or register swapping are examples where this ....

C. A. Knoblock. Automatically Generating Abstractions for Problem Solving. PhD thesis, Carnegie Mellon University, May 1991.

....in fact as dependent on the more immediate features of the environment (such as finding the door out of the room in which it is currently situated) as it is on the eventual goal. Abstraction can be used to guide a planning search and reduce the potentially huge search space to a manageable size [18, 5, 13, 22]. But the work on abstraction assumes that a plan is built without time constraints, so it is not applicable to problems that require incremental results. The work on temporal planning [1, 15, 17, 20] provides formalisms that represent the dynamic and continuous nature of a changing environment ....

....abstract plan. So by arranging the abstract effects as specific goals, the agent will be directed towards the specific plan that best follows the abstract plan. We take the view that the intermediate goals are propagated as an unordered set rather than as a strict sequence of milestones as in [13]. A strict sequence assumes that there is no interaction among the goals, which is unreasonable in real world domains (or even in classic examples such as Sussman s anomaly [21] Using the goals as a set implies that each state is tested against a set of goals and is judged with regard to how ....

[Article contains additional citation context not shown here]

C. A. Knoblock. Automatically Generating Abstractions for Problem Solving. PhD thesis, Carnegie Mellon University, May 1991.

....intuition. From the experimental results, we could expect our study in this paper to contribute to the fields of analogical reasoning and case based reasoning as well as theorem proving. 1 Introduction Using abstraction is an effective approach to improve theorem proving and planning efficiencies [1, 2, 3, 4, 5, 6, 7]. This paper is concerned with an abstraction in theorem proving, especially SLD refutation. 1 Given a (concrete) theory T and a goal G to be proved, theorem proving with abstraction is performed according to the following three steps: Step1 : Abstracting goal The goal G is transformed into a ....

....Since finding similarities is a central task in the field of analogical reasoning and case based reasoning, our study in this paper would make an important contribution to the fields. Knoblock has proposed a framework for constructing an abstraction hierarchy depending on a given planning problem [5]. However the abstraction is formed by eliminating some literals in the problem domain. Therefore we can not find similarities between concepts explicitly. In addition, the experimental results show that the property stated in Theorem 4 26 Appropriate Predicate Mappings for can contain(X) ffl ....

Craig A. Knoblock, "Automatically Generating Abstractions for Problem Solving ", Technical Report CMU-CS-91-120, School of Computer Science, Carnegie Mellon University, 1991.

....a domain model leads to many opportunities to improve the model with respect to its fit with the domain and its efficiency when used at plan time. In the past, classical planning has been concerned with the level of the literal or proposition. Abstraction mechanisms have relied on them (Knoblock [36]) complexity classes rely to some extent on them (Bylander [6, 7] and theoretical formulations rely on them (Chapman [11] Lifschitz [38] The attraction of this is the apparent generality of the approach, yet syntactic restrictions (such as having function free terms or propositional terms) ....

....2. Tools used primarily for compiling a domain model into a more efficient or operational form. These include: a) macro generation (see section 4.1 below) b) generation of various types of goal orderings (see section 4. 2 below) c) abstraction hierarchy generation (for example ALPINE in [35, 36]) At present our environment contains a number of tools of category 1 (a, b, e and f) as follows: ffl a tool that uses the substate class definitions to: check the syntax of the domain operator set; check that all substates that should be achievable are indeed achievable by operator action; help ....

[Article contains additional citation context not shown here]

C. A. Knoblock. Automatically Generating Abstractions for Problem Solving. PhD thesis, School of Computer Science, Carnegie Mellon University, 1991.

.... literals that accounts for the greatest variation in the state utility and use the action representation to find literals that can directly or indirectly affect the chosen set, using a technique similar to the one developed by Knoblock for building abstraction hierarchies for classical planners (Knoblock 1991). This subset of literals then forms the basis for an abstract MDP by projection of the original states. Since the state space size is exponential in the set of literals, this reduction can lead to considerable time savings over the original MDP. Boutilier and Dearden prove bounds on the ....

Knoblock, C. A. 1991. Automatically Generating Abstractions for Problem Solving. Ph.D. Dissertation, Carnegie Mellon University.

....is implementing the abstract plan. So by mapping the abstract effects to specific goals, the agent will construct a specific plan that follows the abstract plan. Much of the existing work on abstraction assumes that intermediate goals are propagated as a strict sequence of goals to be achieved [11, 17, 20]. A strict sequence requires that there is no interaction among the goals, which is unreasonable in real world domains (or even in classic examples such as Sussman s anomaly [19] In contrast, we propagate the intermediate goals as a set, and the specific level will build a plan with respect to ....

C. A. Knoblock. Automatically Generating Abstractions for Problem Solving. PhD thesis, Carnegie Mellon University, May 1991.

....and less expensive search. The space of possible abstractions, based on unmet preconditions for a particular context, is problem dependent. The basis method, as well as other techniques for abstracting preconditions (e.g. ABStrips [22] belongs to the class of Proof Increasing Abstractions [11, 12]. For such abstractions, if a ground solution exists, it is guaranteed that there always exists some abstract solution such that the ground solution may be constructed by monotonically adding steps to the abstract solution. When operator preconditions are abstracted using Spatula s basis method, ....

....in a straightforward manner. 4 Discussion 4.1 Impasse Driven Abstraction There has been much recent research towards the automatic generation of abstractions. Such work has primarily been focused on developing abstractions by performing pre task analyses of a problem solving domain, e.g. [4, 9, 12, 15, 23]. The approach described here does not rely on precomputed analyses, but draws on different sources of knowledge, derived from run time context, to determine the abstractions. Because Spatula s knowledge is obtained during problem solving, it can be applied when pre task analysis does not produce ....

C. Knoblock. Automatically Generating Abstractions for Problem-Solving. PhD thesis, Carnegie-Mellon University, 1991.

....avoided in the further [Dechter, 1989] Hierarchical Reformation Techniques: These are techniques to automatically define abstractions in a CSP. If carefully chosen, such abstractions have been known to significantly reduce search [Stefik, 1981] Fox, 1987] With the exception of [Dechter, 1989][Knoblock, 1991], very little formal work has been done in this area. In this study, we consider a depth first search procedure that starts in a state where all operations (activities) still have to be schedules and proceeds by scheduling operations one by one (see Figure 2) Each time an operation is scheduled, ....

Knoblock, C., Automatically Generating Abstractions for Problem Solving, Ph.D. Thesis, School of Computer Science, Carnegie Mellon University, 1991.

.... They provide a a possible answer to the request that : it should be possible to determine scope from the current context, attention, and goals of the agent : 2] The idea of linking the theory formulation to the problem formulation can be found in some work done in abstraction [1, 12]. In this work, the problem statement is used to drive the contruction of abstract spaces, but not of the ground space, as it happens here. Another interesting similarity can be found with the notion of mental space introduced by Fauconnier [3] The notion of mental space seems somehow similar to ....

C.A. Knoblock. Automatically Generating Abstractions for Problem Solving. PhD thesis, School of Computer Science, 1991.

No context found.

Craig A. Knoblock. Automatically Generating Abstractions for Problem Solving.PhD thesis, School of Computer Science, Carnegie Mellon University, 1991. Available as Technical Report CMU-CS-91-120.

No context found.

Craig A. Knoblock. Automatically Generating Abstractions for Problem Solving. PhD thesis, School of Computer Science, Carnegie Mellon University, 1991. Tech. Report CMU-CS-91-120.

No context found.

Craig A. Knoblock. Automatically Generating Abstractions for Problem Solving. Ph.D. Thesis, School of Computer Science, Carnegie Mellon University, 1991. Tech. Report CMU-CS-91-120.

....problem in an abstract space and then refine the abstract solution at successively more detailed levels in an abstraction hierarchy. This technique has been used successfully to reduce search in a number of planning systems, including gps [48] abstrips [53] abtweak [68] pablo [11] and prodigy [32, 35]. While hierarchical planning is a widely used planning technique, there are only a few systems that automate the construction of abstraction hierarchies [3, 11, 53] In most hierarchical planners, the designer of a planning domain must manually engineer the appropriate abstractions. This process ....

....knowledge. The third subsection compares alpine and abstrips in the original strips domain and shows that alpine produces abstractions that have a considerable performance advantage over those generated by abstrips. The raw data from the experiments described in this section is available in [32]. 5.1 Empirical Results for ALPINE alpine generates abstraction hierarchies for a variety of problem solving domains. This section describes the abstractions generated by alpine on two domains, a robot planning domain and a machine shop planning and scheduling domain, and presents empirical ....

Craig A. Knoblock. Automatically Generating Abstractions for Problem Solving. PhD thesis, School of Computer Science, Carnegie Mellon University, 1991.

No context found.

Knoblock, C.A.: Automatically Generating Abstractions for Problem Solving. PhD thesis, Computer Science Department, Carnegie Mellon University, Pittsburgh, PA (1991)

No context found.

Craig A. Knoblock. Automatically Generating Abstractions for Problem Solving. PhD thesis, School of Computer Science, Carnegie Mellon University, 1991. Technical Report CMU-CS-91-120.

No context found.

Craig A. Knoblock. Automatically Generating Abstractions for Problem Solving. PhD thesis, School of Computer Science, Carnegie Mellon University, 1991. Tech. Report CMU-CS-91-120.

No context found.

Craig A. Knoblock. Automatically Generating Abstractions for Problem Solving. PhD thesis, School of Computer Science, Carnegie Mellon University, May 1991. Tech. Report CMU-CS91 -120.

No context found.

Craig A. Knoblock. Automatically Generating Abstractions for Problem Solving. PhD thesis, School of Computer Science, Carnegie Mellon University, 1991. Tech. Report CMU-CS-91-120.

No context found.

Craig A. Knoblock. Automatically Generating Abstractions for Problem Solving. PhD thesis, School of Computer Science, Carnegie Mellon University, 1991. Tech. Report CMU-CS-91-120.

No context found.

Knoblock, C.A.: Automatically Generating Abstractions for Problem Solving. PhD thesis, Computer Science Department, Carnegie Mellon University, Pittsburgh, PA (1991)

No context found.

Craig A. Knoblock. Automatically Generating Abstractions for Problem Solving. PhD thesis, School of Computer Science, Carnegie Mellon University, Pittsburgh, PA, 1991. Available as technical report CMU-CS-91-120.

No context found.

Knoblock, C.A.: Automatically Generating Abstractions for Problem Solving. PhD thesis, Computer Science Department, Carnegie Mellon University, Pittsburgh, PA (1991)

No context found.

Knoblock, C.A. (1991), Automatically Generating Abstractions for Problem Solving, technical report CMU-CS-91-120, Computer Science Department, Carnegie-Mellon University.

First 50 documents  Next 50

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