Thagard, P., K.J. Holyoak, et al. 1990. Analog Retrieval by Constraint Satisfaction. Tech. Rep. CSL-Report 41, Princeton U.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....have been used for exploiting analogy in problem solving and planning (cf. 1,37] but such systems lack the flexibility to deal with large knowledge bases and cover the range of phenomena that SME and MAC FAC handle. Other cognitive simulations of analogical processing, including ACME and ARCS [31], LISA[25] IAM [26] have only been tested with small examples, and some are known to fail when tested with descriptions even one tenth the size of what was needed to handle the problems described here. No previous analogy systems have been successfully used with multiple, large general purpose ....

Thagard, P., Holyoak, K. J., Nelson, G., & Gochfeld, D. (1990). Analog retrieval by constraint satisfaction. Artificial Intelligence, 46, 259-310.

....will be adapted is ignored; retrieval efficiency is seen as the most important criterion to meet. The implicit assumption in these methods is that the case whose specification is most semantically similar to the target s will also be the most useful case and will require the least adaptation [11,12].These traditional retrieval techniques can produce efficient results, but their assumption that similarity=utility does not always hold; in some situations the efficiency of the retrieval method is traded for gross inefficiency during adaptation. This is because the most similar case to the ....

Thagard, P., Holyoak, K. J., Nelson, G., and Gochfeld, D. Analog Retrieval by Constraint Satisfaction. Artificial Intelligence, 46, 3, pp 259-310, 1990.

....content addressable memories. However, their use in models of analogy processing in humans has been limited by the widespread supposition that it is difficult or impossible to represent compositional structure in vector representations (Fodor and Pylyshyn 1988, Ratcliff and McKoon 1989, Thagard, Holyoak, Nelson and Gochfeld, 1990, Gentner and Markman 1993,Forbus, Gentner and Law, 1994, Wharton, Holyoak Downing, Lange, Wickens, and Melz 1994) This supposition is false. Structure can be represented in vectors in a number of ways, e.g. Smolenskys (1990) tensor products, Pollacks (1990) RAAMs, Kanervas (1996) binary ....

....literature. Much attention has been devoted to teasing apart the differing effects of superficial and structural similarity in retrieval and mapping. A very brief introduction to the concepts and task involved in analogy processing is given here; for more details, see Forbus et al. (1994) Thagard et al. (1990), or Hummel and Holyoak (1997) Analog processing concerns the processes of retrieving stories or scenarios from long term memory in response to some immediate stimulus (another story or scenario) and relating or applying the retrieved story or scenario to the current one. In a typical ....

[Article contains additional citation context not shown here]

THAGARD P, HOLYOAK K J, NELSON G, and GOCHFELD D 1990 Analog Retrieval by Constraint Satisfaction. Artificial Intelligence 46, 259---310.

....one branch. 3. 2 STRUCTURE MATCHING AND STRUCTURE SIMILARITY MEASUREMENT Basically two types of similarity are recognised surface similarity when two problems have the same attributes and structural similarity when they have the same underlying casual dependency network (Holyoak and Koh 1987; Thagard et al. 1994). The proposed algorithm is based on the evaluation of a surface similarity between two design structure spaces or, more accurately, between their design prototypes. The surface similarity can be defined in many different ways. We define it as proportional to the maximal total number of matching ....

Thagard, P., Holyoak, K., Nelson, G. and Gochfield, D.: 1994, Analog retrieval by constraint satisfaction, Artificial Intelligence 46: 259-310.

....selection heuristics as do the Board s reuse of past operationalizations (technique 9) e.g. the case and code instantiations in a cited case can be reused. SIROCCO s Two Stage Retrieval Process Following the general approach in designing analogical retrieval programs (Forbus et al. 1994; Thagard et al. 1990), SIROCCO s retrieval phase is implemented as a two stage algorithm, as shown in Figure 1. Stage 1 rapidly matches the target case s Fact Primitives to those of all possible source cases. Stage 2 applies a more expensive A search to map selected case structures between target and source cases. ....

Thagard, P., K.J. Holyoak, et al. 1990. Analog Retrieval by Constraint Satisfaction. Tech. Rep. CSL-Report 41, Princeton U.

....SIROCCO to match similar cases. In addition, SIROCCO employs a two stage retrieval algorithm intended to match similar cases flexibly enough despite inevitable small variations in the way cases are described. While our two stage approach is based on that of several researchers in analogy (e.g. Thagard et al., 1990; Forbus et al., 1994) we focus more specifically on coverage of time dependent scenarios and leveraging of goal specific knowledge. To date, we have implemented the first stage and report the results of an experiment designed to evaluate it. Case Acquisition and Representation in SIROCCO As ....

Thagard, P., Holyoak, K., Nelson, G., and Gochfeld, (1990). Analog Retrieval by Constraint Satisfaction, Artificial Intelligence 46, Pp. 259-310.

.... 1990) Much of the data on the effects of specific prior problems has been published in the crossdomain analogy literature where baseline performance, without the retrieval and use of a prior problem, is poor (eg Gick Holyoak, 1980) The most complete models of cross domain analogy; ARCS ACME (Thagard, Holyoak, Nelson Gochfield, 1990; Holyoak Thagard, 1989) and MAC FAC SME (Gentner Forbus, 1991; Falkenheimer, Forbus Gentner, 1986) separate the mechanism underlying the use of prior problems into discrete retrieval and mapping stages. A similar mechanism is thought to govern the use of within domain analogy (Reeves ....

Thagard, P., Holyoak, K. J., Nelson, G., & Gochfield, D. (1990). Analog retrieval by constraint satisfaction. Artificial Intelligence, 46, 259-310.

....that explained the data and made some new predictions. Typically, a model of mapping is coupled with a (separate) model of retrieval. The best known examples are SME MAC FAC (Falkenhainer, Forbus, and Gentner, 1986; Forbus, Gentner, and Law, 1995) and ACME ARCS (Holyoak and Thagard, 1989; Thagard, Holyoak, Nelson, and Gochfeld, 1990). However, the experimental work soon revealed that the pattern is not that clear and straightforward. It has been demonstrated that superficial similarities do play an important role in mapping as well. In particular crossmapping is difficult (Ross, 1989) This led Holyoak and Thagard to include ....

Thagard, P., Holyoak, K., Nelson, G., and Gochfeld, D. (1990). Analog retrieval by constraint satisfaction. Artificial Intelligence, 46, 259-310.

....evident that indexed memory will present difficulties in supporting remote or abstract remindings. The work of Waltz and Stanfill and that of Thagard and Holyoak is explicitly directed at cross domain remindings and memory organisation that supports abstract remindings, Thagard Holyoak, 1989; Thagard, Holyoak, Nelson Gochfeld, 1990; Waltz 1989; Stanfill Waltz, 1986) However as a consequence, retrieval is only possible when cases are stored without indexing. This requires that memory is content addressable in that all information about stored cases is matched with the target case and the best match is returned. It is ....

....design using CBR, is that systems designed to operate within one domain may be required to support mappings between sub domains of that domain (see Section 3) Therefore, our perspective is that there exists a complete spectrum of remindings at different levels of abstraction. It has been argued (Thagard, Holyoak, Nelson Gochfeld, 1990; Waltz 1989) that for cross domain remindings, indexing using abstract remindings is problematic and therefore exhaustive search of the case base with subsequent structural mapping is a preferable approach. However it is our opinion that effective EBR systems will only materialise when this ....

Thagard P., Holyoak K.J., Nelson G. & Gochfeld D. (1990). Analog Retrieval by Constraint Satisfaction.

....However, a few simple changes to the programs will fool any similarity metric based on this profile. There are also connectionist techniques for classification and similarity that have not been considered here. In particular spreading activation and constraint satisfaction have been used in ARCS [9] and SAARCS [10] These are localist connectionist systems and are different to other approaches mentioned in this paper. 4 Table 1. Frequency counts of identifiers in two programs Program stu1 Program stu3 19 close 20 close 11 pipea 10 pipea 10 pipeb 10 pipeb 4 hold 4 num 4 dup 4 dup 3 wait ....

Thagard P., Holyoak K.J., Nelson G., Gochfeld D., "Analog Retrieval by Constraint Satisfaction", Artificial Intelligence, pp259-310, Vol. 46, 1990.

....problem. 2. Implementation of a CBR system on a parallel computer without using case classification (Stanfill and Waltz, 1986; Waltz, 1989; Rubin, 1992) 3. Classification is supported by parallel implementation (Kolodner, 1989) 4. Neither classification nor the parallel implementation is used (Thagard et al. 1990; Hennessy and Hinkle, 1991; Gentner and Forbus, 1991) However, this approach is only usable for small case bases. 4.1 Classification Methods Different systems use different classification methods. There are several possibilities to categorize these methods. Porter (Porter, 1989) categorize the ....

....features (superficial, perceivable, intrinsic) on the basis of abstract features (derived, deep) or based on both. In many studies (e.g. Tversky, 1977; Gentner, 1983; Birnbaum and Collins, 1988; Keane, 1988; Vosniadou and Ortony, 1989; Collins and Birnbaum, 1990; Simoudis and Miller, 1990; Thagard et al. 1990; Gentner and Forbus, 1991; McDougal, Hammond and Seifert, 1991; Kolodner, 1992; Owens, 1993) researchers are aware of the existence of all three types of remindings; however, there is a disagreement on what features are the most useful and important for retrieval (these features are referred to ....

[Article contains additional citation context not shown here]

Thagard, P. R., Holyoak, K. J., Nelson, G., and Gotchfeld, D. (1990). Analog retrieval by constraint satisfaction. Artificial Intelligence, 46:259--310.

....not only exact but similar matches as well. We say that a case satisfies a particular context, if for each attribute specified in the context, the value of that attribute in the case satisfies the constraint. Thus, the matching process can be described as a constraint satisfaction process [THNG90,Gaa93]. One of the resulting problems is selecting a method for specifying a context, i.e. specifying which attributes are needed for accurate classification, and what range or similarity should be allowed for attribute values (i.e. what constraints should be posed on values) The system should ....

Thagard, P. R., Holyoak, K. J., Nelson, G., and Gotchfeld, D. Analog retrieval by constraint satisfaction. Artificial Intelligence , 46:259-310, 1990.

....[Hof84] Dioeerent computational models have been developed to model analogy solving. Among them, the ANALOGY system proposed by Evans [Eva68, Ha91] the SME system proposed by Falkenhainer to illustrate Gentner s theory for analogy [FFG89, Gen83] the ARCS system developed by Thagard and Holyoak [THNG90] to simultaneously satisfy the structural, semantic and pragmatic constraints. We can hardly compare these systems to ESQIMO in terms of performances since we only studied intra domain analogies with the only structural constraint in this rst work. Our contribution lies principally in the search ....

....this rst work. Our contribution lies principally in the search for a new representational structure to model analogy in terms of diagrams. The CAT structure of representation can be seen as a hybrid structure between a purely symbolic and a purely analogical approach, like neural networks in ARCS [THNG90]. ESQIMO has been prototyped in the Mathematica programming language and implemented in ML to include many of the heuristics discussed before. Even if this model can be considered as very simple, we are convinced that a topological representational structure is well adapted for analogy modeling. ....

P. Thagard, K. J. Holyak, G. Nelson, and D. Gochfeld. Analog retrieval by constraint satisfaction. Artiøcial Intelligence, 46(3):259310, December 1990.

....might suggest an old proof, yet these are ignored in Carbonell s scheme unless they also suggest the initial steps in the solution. Finally, Carbonell provides few details on his retrieval ideas. 6. 3 ARCS The ACME analogical mapping system is accompanied by a retrieval system called ARCS [20]. The basic ideas in ARCS are the same as those in ACME with only a few differences. Like ACME, ARCS is based on semantic, structural, and pragmatic constraints, and it uses a parallel, competitive relaxation of these constraints. However, unlike ACME, semantic similarity is the dominant pressure ....

....Like ACME, ARCS is based on semantic, structural, and pragmatic constraints, and it uses a parallel, competitive relaxation of these constraints. However, unlike ACME, semantic similarity is the dominant pressure in ARCS, due mostly to the psychological evidence for such an approach in humans [20]. ARCS uses three knowledge representations: Predicate Calculus for the source and target analogs, frames attached to predicates for semantic information, and a constraint network for analogical correspondences. Each semantic frame includes a set of synonyms, antonyms, tenses, plurals, ....

[Article contains additional citation context not shown here]

Thagard, P, K. Holyoak, G. Nelson, and D. Gochfeld, "Analog Retrieval by Constraint Satisfaction" unpublished manuscript, Cognitive Science Laboratory, Princeton University, Princeton, NJ, February 1989.

....certain extent. In the recent years it has become clear that all three constraints are important at least at some steps in the reasoning process. So Gentner (1989) included pragmatic constraints in her reasoning model (although only external to the mapping engine) while Holyoak Thagard (1989b) Thagard, Holyoak, Nelson Gochfeld (1990) included structural constraints both on mapping and retrieval and built the ACME and ARCS models governed by all three types of constraints. Holyoak Thagard (1989b) give clear definitions of structural, semantic and pragmatic constraints. A structural constraint is the pressure to find and use ....

....with respect to a particular context, hence two questions arise: what is considered as a context and what are the criteria for determining the relevance. Typically only the problem context is taken into consideration, i.e. the relevance of an element is defined with respect to the problem goal (Thagard et al. 1990) or, at most, with respect to the whole problem description (Anderson, 1983, Mitchell Hofstadter, 1990) In some models (Eskridge, this volume) the contextual goal of the reasoner (e.g. problem solving, learning, explaining, etc. is also taken into account. I would like, however, to consider ....

[Article contains additional citation context not shown here]

Thagard, P., Holyoak, K., Nelson, G., Gochfeld, D. (1990). Analog Retrieval by Constraint Satisfaction. Artificial Intelligence, vol. 46, pp. 259-310.

....should support more complex mechanisms in models of different processes. The cognitive architecture should ensure context sensitivity of all cognitive processes. Reviewing the existing models of analogy making and especially those of them that involve reminding of an old episode ARCS (Thagard, Holyoak, Nelson, Gochfeld, 1990), MAC FAC (Forbus, Gentner, Law, 1995) AMBR1 (Kokinov, 1994a) and LISA (Hummel Holyoak, 1997) we will notice that they fail to incorporate most of the behavioral and architectural Kokinov Petrov Memory and Reasoning in AMBR 20 constraints described here. 3 Invariably these models use ....

Thagard, P., Holyoak, K., Nelson, G., & Gochfeld, D. (1990). Analog retrieval by constraint satisfaction. Artificial Intelligence, 46, 259-310.

No context found.

Thagard, P., K.J. Holyoak, et al. 1990. Analog Retrieval by Constraint Satisfaction. Tech. Rep. CSL-Report 41, Princeton U.

No context found.

Paul Thagard, Keith J. Holoyak, Greg Nelson, and David Gochfeld. 1990. Analog retrieval by constraint satisfaction. Artificial Intelligence, 46(3):259--310.

No context found.

Thagard, P., K.J. Holyoak, et al. 1990. Analog Retrieval by Constraint Satisfaction. Tech. Rep. CSL-Report 41, Princeton U.

No context found.

Thagard P., Holyoak K.J., Nelson G., Gochfeld D., (1990), "Analog Retrieval by Constraint Satisfaction" in Artificial Intelligence, Vol. 46, pp259-310, 1990.

No context found.

Thagard, P., Holyoak, K.J., Nelson, G., and Gochfeld, D. (1990). Analog Retrieval by Constraint Satisfaction. In Artificial Intelligence, volume 46, pp. 259-310.

No context found.

Thagard, P., Holyoak, K.J., Nelson, G., and Gochfeld, D. (1990). Analog Retrieval by Constraint Satisfaction. In Artificial Intelligence, volume 46, pp. 259-310.

No context found.

P. Thagard, K. Holyoak, G. Nelson, D. Gochfeld, Analog Retrieval by Constraint Satisfaction. Artificial Intelligence, 46, (1990).pp. 259-310.

No context found.

Thagard P. et al. 1990. Analog Retrieval by Constraint Satisfaction, Artificial Intelligence, 46

No context found.

Thagard, P., K. J. Holyoak, G. Nelson, and D. Gochfeld (1990). Analog Retrieval by Constraint Satisfaction. Artificial Intelligence 46, 259--310.

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