Hayes-Roth F., McDermott J. (1978) "An interference matching technique for inducing abstractions"Comm. ACM. Artificial Intelligence, Language processing.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....negation as failure) the meaning of each example in the above representation is correctly captured (i.e. if color left(black) then color left(X) will only succeed for X=black. This framework and use of Prolog is quite similar to what happens in the older work on structural matching (e.g. [38, 39, 70, 71, 72]) 2.3 Background knowledge It is useful to use not only factual knowledge in the examples, but also Prolog rules (or definite clauses) If these rules are common to all the examples, they are referred to as background knowledge. Such knowledge can take various forms: e.g. abstraction of ....

F. Hayes-Roth and J. McDermott. An interference matching technique for inducing abstractions. Communications of the ACM, 21:401--410, 1978.

....be employed to analyze software resource data. The application of the above framework to a number of existing systems is presented in Figure 2 to better understand possible alternative methods (see also [DM79] and [DM83] The systems characterized in the figure are Sprouter[Hay75] Hay76] HM77] [HM78], Induce[DM79] Arch[Win75] Win77] Lex[Mit82] MUNB83] and ID3[Qui85] Systems examined in the figure reflect by no means the entirety of such systems, but are intended to be a representative sampling. As explained in Section 3.1.1, the ID3 system was selected as the basis for the tools ....

F. Hayes-Roth and J. McDermott. An interference matching technique for inducing abstractions. Communications of the ACM, 21(5):401--410, 1978.

....of terms in the clause. So indeed it is a polynomial transformation. 6 2.3 The Graph Matching Approach of Learning MSGs Starting with Winston [26] the other main line of computing MSGs was established. Other approaches in this line of research are Hayes Roth s and McDermott s system SPROUTER [6, 7] and the work of Vere [22, 23, 24] A formal description and a proof of the non PAC learnability of these MSG approaches is done by Haussler [5] This framework of learning MSGs forms the other main line of learning MSGs. Whereas Plotkin and Helft regard learning from a logical point of view, ....

Frederick Hayes-Roth and John McDermott. An interference matching technique for inducing abstractions. CACM, 21:401--410, 1978.

....arcs and not the largest common sub graphs. In doing so, it ignores the fact that any two arcs may be connected. In other words, a graph is considered as a set of arcs; matching graphs is viewed as matching arcs. It consists in restricting the generalization language to a one to one matching (Hayes Roth and McDermott, 1978). As a result of this restriction, the complexity of the graph matching process is polynomial. Nevertheless, as previously explained in 2.2.3. the GS does not contain all the existing generalizations of objects (only the ones that may be expressed in this restricted language) 7 3.2. Size of ....

Hayes-Roth, F. and McDermott, J. (1978). An interference matching technique for inducing abstractions. Communications of ACM, 21(5), pp.401-410.

....instance space defined by n attributes is NP complete. Furthermore, he shows that the size of MSCG can grow exponentially with the number of examples m. Although Haussler notes that heuristic methods for learning conjunctive concepts can be effective, the existing inductive algorithms (Vere 1980; Hayes Roth and McDermott 1978; Watanabe and Rendell 1990) do not apply well to our operator learning problem. For example, Vere s counterfactual algorithm (Vere 1980) requires both positive and negative examples and is not incremental. Hayes Roth s interference matching algorithm (Hayes Roth and McDermott 1978) uses some ....

....(Vere 1980; Hayes Roth and McDermott 1978; Watanabe and Rendell 1990) do not apply well to our operator learning problem. For example, Vere s counterfactual algorithm (Vere 1980) requires both positive and negative examples and is not incremental. Hayes Roth s interference matching algorithm (Hayes Roth and McDermott 1978) uses some complex heuristics to prune the space of all the matches between two examples, and thus prevent some operator preconditions from being learned. It also uses the one to one parameter binding assumption that does not hold for our operator preconditions ( in general, two or more different ....

F. Hayes-Roth and J. McDermott. An interference matching technique for inducing abstractions. In CACM, volume 26, pages 401--410, 1978.

....are important for their contribution to the understanding of induction in structural domains. For each system, we discuss its representation language for instances and concepts, its classification mechanism, and its learning algorithm. 2. 1 Sprouter: Incremental Learning with Structured Objects Hayes Roth and McDermott s (1978) Sprouter is representative of several systems that carry out learning of maximally specific conjunctive descriptions from examples (e.g. Vere, 1975; Winston, 1975) These systems focus on finding characterizations, or descriptions, of classes given by an external teacher. Dietterich and ....

....an initial partial match; it then selects a new case frame from I and repeats the process. If the bindings between the frames are consistent with the previous 1. This case frame representation appears to play a role similar to that of the attributes used by many inductive learning systems. Hayes Roth and McDermott (1978, p. 402) use this idea of defining certain shared properties. In addition, SPROUTER appears to use the case frames to direct matching between properties that other systems would represent as n ary predicates (e.g. ABOVE, BELOW) 2. In contrast, Vere s THOTH (1975) considers all maximal ....

[Article contains additional citation context not shown here]

Hayes-Roth, F., & McDermott, J. (1978). An interference matching technique for inducing abstractions. Communications of the ACM , 21 , 401--410.

....an instance space defined by n attributes is NPcomplete. Furthermore, he shows that the size of MSCG can grow exponentiallywith the number of examples m. Although Haussler notes that heuristic methods for learning conjunctiveconcepts can be effective, the existing inductive algorithms [Vere, 1980, Hayes Roth and McDermott, 1978, Watanabe and Rendell, 1990] do not apply well to our operator learning problem. For example, Vere s counterfactual algorithm [Vere, 1980] requires both positive and negative examples and is not incremental. Hayes Roth s interference matching algorithm [Hayes Roth and McDermott, 1978] uses some ....

....[Vere, 1980, Hayes Roth and McDermott, 1978, Watanabe and Rendell, 1990] do not apply well to our operator learning problem. For example, Vere s counterfactual algorithm [Vere, 1980] requires both positive and negative examples and is not incremental. Hayes Roth s interference matching algorithm [Hayes Roth and McDermott, 1978] uses some complex heuristics to prune the space of all the matches between two examples, but it may prevent some operator preconditions from being learned. It also uses the one to one parameter binding assumption that does not hold for our operator preconditions ( in general, two or more ....

F. Hayes-Roth and J. McDermott. An interference matching technique for inducing abstractions. In CACM, volume 26, pages 401-- 410, 1978.

....the multistrategy integration, the results obtained by the systems in some application domains are summarized. 1. Introduction Learning knowledge expressed in First Order Logic (FOL) has been an appealing task since the beginning of Machine Learning. Early attempts (Plotkin, 1970; Winston, 1975; Hayes Roth and McDermott, 1978; Vere, 1978; Michalski, 1980; Dietterich and Michalski, 1983; Kodratoff and Ganascia, 1986) have proposed a number of conceptually interesting ideas, which provided foundations and suggestions for later work. However, the need of excessive computational resources made the proposed methods rather ....

Hayes-Roth, F. and McDermott, J. (1978). An interference matching technique for inducing abstractions. Communications of the ACM, 21:401--411.

No context found.

Hayes-Roth F., McDermott J. (1978) "An interference matching technique for inducing abstractions"Comm. ACM. Artificial Intelligence, Language processing.

No context found.

Hayes-Roth F., McDermott J., 1978, "An interference Matching Technique for inducing Abstractions, CACM, pp. 401-411.

No context found.

Hayes-Roth, F., & McDermott, J. (1978). An Interference Matching Technique for Inducing Abstractions. Communications of the ACM, 21, 401-411.

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