George W. Ernst and Allen Newell. GPS: A Case Study in Generality and Problem Solving. ACM Monograph Series. Academic Press, New York, NY, 1969.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....to integration: instead of trying to find one general method for all cases (which we do not believe is possible anyway [16] 7] we combine general methods for certain processes with powerful specialized methods that apply in certain contexts. For example, the means end analysis of GPS [63] [21] [64] is a general method for searching through a space, but it needs to be augmented with special searches organized differently for different spaces [4] We have divided the Study problem process into three main steps: Interpret problem , which means to find a resource to apply to the ....

G. Ernst, A. Newell, GPS: A Case Study in Generality and Problem Solving, ACM Monograph Series, Academic Press (1969)

....by attempting to reduce the differences between the initial state and goal. The problem of finding good orderings of the differences has been extensively explored in gps and is closely related to the techniques for generating abstractions in alpine. The criterion for ordering the differences in [12, 15] is to attempt to find an ordering such that achieving one difference will not affect a difference reduced by operators selected earlier in the ordering. The algorithm for finding an ordering requires building a table of differences and finding a lowertriangular difference table. This is similar ....

George W. Ernst and Allen Newell. GPS: A Case Study in Generality and Problem Solving. ACM Monograph Series. Academic Press, New York, NY, 1969.

....(TYPE X) DESIRED VALUE: LIST, RELATED WORK 86 Constraint 2 DESIRED PROP: LENGTH X) DESIRED VALUE: 3, and Constraint 3 DESIRED PROP: DEPTH (FIRST ELT X) DESIRED VALUE: 1. The major knowledge base of CEG consists of an examples space and difference operator table similar to that of GPS [15]. The examples space consists of known examples LISP data structures organized according to the relation of constructional derivation representing which examples are constructed from which others. The CEG example generation process consists of three major phases: retrieval of known examples, ....

Ernst, George W. and Newell, Allen. GPS: A Case Study in Generality and Problem Solving. Academic Press, New York, 1969.

....of existing knowledge during the problem solving process has been systematically ignored in the problem solving literature. Unfortunately, most classical models of problem solving take the existing knowledge as constant 1 . Long term memory structures only have to be retrieved and applied (Ernst Newell, 1969; Newell, 1990; Anderson, 1983; Anderson Lebiere, 1998) The same is true for models of analogy making: they retrieve a ready made representation of an old episode when looking for a base for analogy (Gentner, 1989; Thagard, Holyoak, Nelson Gochfeld, 1990; Kokinov, 1994a; Forbus, Gentner ....

Ernst, G. & Newell, A. (1969). GPS: A case study in generality and problem solving. NY: Academic Press.

.... to literature such as [Drummond Tate 87] The most influential single planning system is probably Strips, produced in 1972 [Fikes et al. 72] This is a state space planner that applies the principle of meansends analysis (MEA) first introduced by the General Problem Solver (GPS) project [Ernst Newell 69] MEA is a search heuristic which works by selecting operators that can be shown to be significant in reducing the differences between the current state in the search space and the desired state. However, differences are reduced one by one, which means that the heuristic will not always be able ....

G. Ernst and A. Newell. GPS: a case study in generality and problem solving. Academic Press, New York, 1969.

....in AI comprises methods for finding solutions that are independent of the problem domain. Allen Newell, Herbert Simon and their colleagues and students pioneered this approach and continue to pursue it. Newell and Simon first proposed the General problem Solver GPS in their (1957) also see (Ernst and Newell 1969). The initial idea was to represent problems of some general class as problems of transforming one expression into another by means of a set of allowed rules. It was even suggested in their (1960) that improving GPS could be thought of as a problem of this kind. In my opinion, GPS was unsuccessful ....

Ernst, George W. and Allen Newell (1969). GPS: A Case Study in Generality and Problem Solving, Academic Press.

.... AI research concentrated Chapter 2: Everyday Activities 12 largely on activities that fall into the latter group: those that normally require great concentration in humans, and are at the limits of human mental capabilities [Wilensky, 1983] For example, the General Problem Solver (GPS) [Ernst and Newell, 1969], a direct ancestor of most modern planning systems, was based on a series of studies of human reasoning in solving logic, chess, and cryptarythmetic problems [Newell and Simon, 1972] The types of problems that these systems dealt with have come to be known as microworlds [Drefus, 1981] since ....

....problems [Fikes and Nilsson, 1972] The first such system, GPS (the 29 Nilsson [1971] provides an overview of many behaviours that fall into this category, and common techniques AI uses to replicate them. Chapter 3: AI Planning Models and Everyday Activities 52 General Problem Solver) [Ernst and Newell, 1969], was based on studies of human cognition in solving logic, cryptarithmetic, and chess problems [Newell and Simon, 1972] While GPS nicely modelled the type of reasoning used in these situations, the search space U in the types of problems studied was always reasonably small. For any significant ....

Ernst, George, and Allen Newell, GPS: A Case Study in Generality and Problem-Solving (New York: Academic Press), 1969. 294 pp.

....Corpus 1 3 Errs = 711 Corpus 2 1 Errs = 289 Corpus 2 2 Errs = 341 Figure 4.2: Learning Transformations Corpus 1 Corpus 2 T1 T2 T3 Corpus 0 Figure 4.3: Applying Transformations Transformation based error driven learning is a degenerate instance of means ends analysis. GPS (General Problem Solver) [41, 86] is probably the earliest successful implementation of a means ends analysis system. In GPS, a set of rules is specified. Rules have two parts: the preconditions that must be satisfied to trigger a rule, and the effect of carrying out the rule. The search strategy employed in GPS is more complex ....

G. Ernst and A. Newell. GPS: A case study in generality and problem solving. Academic Press, 1969.

....than the original implementation, while preserving the original flavor. Chapter 4 An Approach to ECM Planning 4.1 Progressive Horizon Planning 4.1. 1 Planning as Search in a Situation Action Space Classical planners originated from theorem provers and search based problem solvers such as GPS [63, 111, 43]. There are several ways in which planning can be viewed as a search task: Classical planners (e.g. Strips [44] Noah [137] Sipe [159] Hacker [144] Nonlin [150] view planning as a search in a space of plans. In a plan space, states represent plan structures and transitions correspond to ....

....hardly useful in an algorithmic sense. Given the complexity of planning as search, in both the plans space and the situation action space [14] much of the research in that area has focused on ways to reduce search. Some common tactics includes: ffl Use of heuristics in focusing the search, e.g. [66, 43]; ffl Resource bounded search, e.g. 89, 91, 134] ffl Means ends analysis, e.g. 44] ffl Search for partially ordered plans, e.g. 137, 150, 160] ffl Least commitment approach, e.g. 142, 74] ffl Abstraction of plan and operator descriptions, e.g. 136, 85, 152, 46] ffl Defaults and ....

Ernst, G., and Newell, A., GPS: A Case Study in Generality and Problem Solving. Academic Press, New York, 1969.

....intelligence. Thus the main emphasis in AI research has been on the finding of general purpose methodologies and general purpose representations that preserve as much information as possible. Such efforts led to the rapid development of various programs such as the General Problem Solver (GPS)[Ernst and Newell, 1969] which was intended to solve a large variety of problems; MACSYMA [Moses, 1976] a program to solve numerous types of mathematical problems; and a geometry program [Gelernter, 1959] which its creator claimed to perform better than himself. A machine that possesses artificial intelligence must be ....

G.W. Ernst and A. Newell. GPS: A Case Study in Generality and Problem Solving. Academic Press, New York, 1969.

....example, G0 can be proved directly from M4, hence the solution sequence of operators (i.e. the plan) is ( OP c, OP b, OP e ) ffl Problem solving through state space search, like the means end analysis (see section 2. 2) and problem reduction performed by the GPS (Generality and Problem Solving) [1] system, built in the late 1950 s by Newell, Shaw and Simon. While STRIPS uses most of the control architecture of GPS, it can handle more complex and general models than GPS can. ffl Green s theorem proving problem solver [4] and situation calculus. STRIPS provides more powerful search heuristics ....

G. Ernst and A. Newell. GPS: a Case Study in Generality and Problem Solving. ACM Monograph Series, Academic Press, New York, 1969.

....treated in a purely symbolic way using the results of the syntactic analysis. This is not surprising, since Computer Vision was considered as a subfield of Artificial Intelligence and thus studied using the same methodology, influenced by the ideas and computational theories of the last decades [12, 21, 44]. The strict hierarchical organization of representational steps in the Marr paradigm makes the development of learning, adaptation and generalization processes practically impossible (so that there hasn t been much work on vision and learning ) Furthermore, the conceptualization of a vision ....

G. Ernst and A. Newell. GPS: A Case Study in Generality and Problem Solving. Academic Press, New York, 1969.

....of the DENDRAL experiments were being presented against that background and as a contrast to the mainstream. The title of the paper was chosen to signal that to the reader. It was an echo of the title of a well known book of its time, the book by Ernst and Newell on generality and problem solving (Ernst and Newell, 1969) Our major experimental generalization was presented and defended: that in a program s knowledge lies its power; that to achieve high levels of competence in performance, AI programs must be knowledge intensive. This theme was elaborated in my 1977 IJCAI invited paper, bringing together the ....

Ernst, G. and Newell, A., GPS: A Case Study in Generality and Problem Solving, New York: Academic Press, 1969.

....This led us to develop our Progressive Horizon planning paradigm as a way of addressing those features. To understand the differences between TraumAID s planner and the classical ones, it is important to remember that the latter (e.g. Strips, 8] descend from general problem solvers (e.g. GPS [7]) that themselves originate from theorem provers and search programs. These planning programs inherited many of the assumptions of their predecessors: they viewed planning as an independent process that takes a goal and an initial world state as its input and produces a sequence of actions (plan) ....

G. Ernst, and A. Newell, GPS: A Case Study in Generality and Problem Solving. Academic Press, New York, 1969.

....of AND OR goal trees constructed by a collection of antecedent and consequent pattern invoked programs driven by an automatic chronological backtracking control structure. See Winston [1976] for discussions of these techniques. In this style of problem solving, embodied in programs like gps [Ernst and Newell 1969], saint [Slagle 1963] Gelernter s [1963] Geometry Theorem proving machine, Black s [1968] question answerer, strips [Fikes and Nilsson 1971] shrdlu [Winograd 1972] and Goldstein s [1973] geometry theorem prover, search through different branches of the goal tree have essentially no influence on ....

George W. Ernst and Allen Newell, GPS: A Case Study in Generality and ProblemSolving, Academic Press, New York, 1969.

....a framework to investigate the mapping from planning domains and problems to efficient planning strategies. 1. Introduction General purpose planning has a long history of research in Artificial Intelligence. Several different planning algorithms have been developed ranging from the pioneering GPS (Ernst Newell, 1969) to a variety of recent algorithms in the SNLP (McAllester Rosenblitt, 1991) family. At the most basic level, the purpose of planning is to find a sequence of actions that change an initial state into a state that satisfies a goal statement. Planners use the actions provided in their domain ....

Ernst, G. W., & Newell, A. (1969). GPS: A Case Study in Generality and Problem Solving.

....planners often search through thousands of plans to solve seemingly simple problems. The alternative search space I describe is much closer to the space searched by the Prodigy planner [36, 34] in that it is based on means ends analysis, a classic search technique first embodied in the GPS system [26, 10]. The principal difference is that Prodigy, like GPS and Strips [13] uses as (define (addendum adjacent def) domain grid) axiom :vars ( i j i1 integer) implies (adjacent i j i1 j right) context (and (equation ( i 1) i1) legalcoord i) legalcoord i1) axiom :vars ( i j ....

....it never undoes this decision. 0 5 10 15 0 1 2 3 4 5 6 7 8 9 Problem Size Time (sec) Times are in tenths of a second Figure 8: Graph of Results for D 1 S 1 7 Relation to Previous Work The present work derives from an attempt to simplify the GPS control structure, which, as described by [10], is rather arcane and complex. The idea of cashing out all matching operations to generate a graph structure linking top level goals to feasible actions first appeared in [7] Section 5.7, where the phrase operator difference tree was used for what I now call the regression match graph. ....

George W. Ernst and Allen Newell. GPS: A Case Study in Generality and Problem Solving. Academic Press, 1969.

....are otherwise intractable. Keywords: planning, search, means ends analysis Reinventing GPS Means ends analysis is one of the oldest ideas in AI. It was named and studied by Newell, Shaw, and Simon in the 1950s, and was the key idea behind the General Problem Solver (GPS) Newell Simon 1961; Ernst Newell 1969) In the late sixties, Fikes, Nilsson, and Raphael embodied the idea in their planner, Strips (Fikes Nilsson 1971) It is still an important technique today, especially as embodied the Prodigy planner (Fink Veloso 1994) As used by planners, means ends analysis can be described thus: We are ....

Ernst, G. W., and Newell, A. 1969. GPS: A Case Study in Generality and Problem Solving. Academic Press.

No context found.

George W. Ernst and Allen Newell. GPS: A Case Study in Generality and Problem Solving. ACM Monograph Series. Academic Press, New York, NY, 1969.

No context found.

Ernst, G. W., and Newell, A., 1969. GPS: A Case Study in Generality and Problem Solving, New York: Academic Press.

No context found.

Ernst, G., and Newell, A., "GPS: A Case Study in Generality and Problem Solving," Academic Press, New York, 1969.

No context found.

Ernst, G. & Newell, A. (1969) GPS: A Case Study in Generality and Problem Solving, ACM Monograph Series, Academic Press, New York.

No context found.

G. W. Ernst and A. Newell. GPS:A case Study in Generality and Problem Solving. Academic Press, New York, 1969.

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