Jaime G. Carbonell. Learning by analogy: Formulating and generalizing plans from past experience. In Ryszard S. Michalski, Jaime G. Carbonell, and Tom M. Mitchell, editors, Machine Learning: An Artificial Intelligence Approach, pages 371--392. Tioga Publishers, Palo Alto, ca, 1983.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....patching failed plans. 1 Introduction Theorem proving by analogy is a process in which the experience of proving a source theorem guides the search for a proof of a similar target theorem. There are at least two different paradigms for analogy: transformational analogy and derivational analogy [2]. Transformational analogy maps a final source solution to the target and may involve additional transformations so the constraints of the target problem are satisfied. Derivational analogy instead maps the problem solving decisions made in the source case [3] and uses these mapped decisions as a ....

J.G. Carbonell. Learning by analogy: Formulating and generalizing plans from past experience. In R.S. Michalsky, J.G. Carbonell, and T.M. Mitchell, editors, Machine Learning: An Artificial Intelligence Approach, pages 371--392. TiogaPubl., Palo Alto, 1983.

....ffl induction chooses induction variables and an appropriate induction scheme. Its output are the subgoals for the base cases and the step case. 3 Analogy Driven Proof Plan Construction in CL A M In Artificial Intelligence, two different paradigms for analogy are: transformational analogy [5] and derivational analogy [6] Transformational analogy maps a final source solution to the target and may involve additional transformations so the constraints of the target problem are satisfied. Derivational analogy instead maps the problem solving decisions made in the source case and uses ....

J.G. Carbonell. Learning by analogy: Formulating and generalizing plans from past experience. In R.S. Michalsky, J.G. Carbonell, and T.M. Mitchell, editors, Machine Learning: An Artificial Intelligence Approach, pages 137-- 162. Tioga, Palo Alto, 1983.

....good reviews of this taxonomy and [Shavlik and Dietterich 90] contains many of the key papers discussing various facets of machine learning. PDL and Machine Learning Machine learning can be divided into several categories, typically including inductive learning [Michalski 83] learning by analogy [Carbonell 83] speedup learning [Mitchell et al. 83] and unsupervised learning [Rumelhart and Zipser 85] Inductive learning is characterized by the acquisition of knowledge by inductive inferences from examples. Learning by analogy consists of using similarity with known facts or procedures to acquire or ....

Carbonell, J.G. Learning by Analogy: Formulating and Generalizing Plans from Past Experience. In Michalski, R.S., Carbonell, J.G., and Mitchell, T.M. (Eds.), Machine Learning: An Artificial Intelligence Approach. Morgan Kaufmann Publishers, Inc., Vol. I, 1983, Chapter 5.

....first find the right level of abstraction. The paradigm that Melis follows is that of derivational analogy. This idea is described by Carbonell, in [Carbonell 85] Carbonell had previously implemented a transformational analogy in ARIES (Analogical Reasoning and Inductive Experimentation System) Carbonell 83] Transformational analogy is based entirely on a comparison between the problem representations, without regard to the decisions made in finding the solutions. The target solution is obtained by adjusting or tweaking the source solution. These heuristic tweaks are motivated by differences in the ....

J.G. Carbonell. Learning by analogy: Formulating and generalizing plans from past experience. Machine Learning, An Artificial Intelligence Approach, R.S. Michalski, J.G.Carbonell, T.M. Mitchell, eds. Tioga Press, Palo Alto, CA, 1983.

....Analogical reasoning typically uses similarity with known situations to make inferences about new situations. For example, in the context of problem solving, one may use plans about previously solved problems to derive, via a series of transformations, a plan for a new, but similar, problem [CAR83]. The identification of similarities between situations varies from simple feature overlap [SUN92] to more complex theories involving both features and the relationships between them [GEN89] An application of analogical learning and reasoning to Szechwan cooking is in [HAM89] One particularly ....

Carbonell, J.G. (1983). Learning by Analogy: Formulating and Generalizing Plans from Past Experience. In Michalski, R.S., Carbonell, J.G., and Mitchell, T.M., (Eds.). Machine Learning: An Artificial Intelligence Approach. Tioga Publishing Company, Palo Alto, CA, Chapter 5.

....scenario generator must now complete a scenario analogous to JOB DAYDREAM in the NUART EPISODE context. Analogous details are filled in using the same planning knowledge that the scenario generator would need to generate NUART DAYDREAM5 from scratch. However, in this case, less effort is required (Carbonell, 1983). The use of recalled episodes is not limited to the generation of entire daydream sequences. Episodic memory is also useful for suggesting possible continuations, or next events, of an ongoing scenario. For example, when DAYDREAMER imagines he is going out with a movie star in NUART DAYDREAM3, he ....

Carbonell, J. G. (1983). Learning by analogy: Formulating and generalizing plans from past experience.

....model rather than to differences in the contexts (they consider only the problem description) In the following subsections I will review the models proposed for retrieval and mapping in more detail. 7.2.1. Retrieval. There are several main approaches to retrieval. 1) In case based reasoning (Carbonell, 1983, 1986, Kolodner and Simpson, 1986, 1989, Hammond, 1989, 1990) retrieval is performed on the basis of a specific organization of LTM around an indexing scheme. Thus Carbonell (1986) indexes the potential sources of analogy by the first reasoning steps in the problem solving activity of that case ....

Carbonell, J. (1983). Learning by Analogy: Formulating and Generalizing Plans from Past Experience. In: Michalski, Carbonell, Mitchell (eds.) Mach. Learning, Palo Alto, CA: Tioga Publ. Comp.

....TH 0 2 : TH 0 n ) and try to use it to prove TH 0 . If that works immediately, then fine. But, for interesting problems it will fail, and the plan has to be changed. How can we 4 debug the plan (TH 0 1 TH 0 2 : TH 0 n ) to make it work The literature on Analogy, for example [1, 20, 26, 18, 13, 10, 14], and many others, has addressed this problem with some degree of success. Here we show how the PC prover can be used to do this debugging for a certain class of theorems including some related to GCR. The basic idea is that PC, in trying to prove the step (TH 0 1 TH 0 2 : TH 0 i ....

Carbonell, J. G., "Learning by Analogy: Formulating and Generalizing Plans from Past Experience", Machine Learning, Michalski, Carbonell, Mitchell (eds.), Tioga Publishers, 1983, 137-161.

....propose to correct its failures by an abductive recovery mechanism inspired from abductive recovery from failed proofs. 1 Introduction The analogy scheme we shall use in this paper is quite a classical one (Winston, 1982; Gentner, 1983; Chouraqui, 1985; Falkenhainer, Forbus, and Gentner, 1986; Carbonell, 1983, 1986; Kedar Cabelli, 1988; Kodratoff, 1988) It can be described as follows. Let us suppose that we dispose of a piece of information, the base, that can be put into the form of a doublet (A, B) in which it is known that B depends on A. This dependency will often be causal, and it does not need ....

Carbonell, J.G. Learning by Analogy: Formulating and Generalizing Plans from Past Experience, in R.S. Michalski, J. G. Carbonell, T. M. Mitchell (Eds.), Machine Learning: An Artificial Intelligence Approach, Morgan Kaufmann 1983, pp. 137-159.

....Furthermore, since similarities might be found at a different level from the purely superficial one of entities observable characteristics, the entities are often analyzed using deductive, or explanation based, mechanisms as well. Examples of work in analogy include [Burstein 86; Carbonell 83a; Carbonell 83b; Kedar Cabelli 85; Russell 86] Exemplar based learning is another mechanism that combines inductive and deductive techniques. In exemplar based learning a concept is represented by a stereotypical example. Deduction, or explanation derivation, is used to determine whether an input might be an ....

Carbonell, J. C. Learning by Analogy: Formulating and Generalizing Plans from Past Experience. Machine Learning: An Artificial Intelligence Approach. Morgan Kaufmann, Los Altos, California, 1983.

....in the planning process. 1 Introduction Theorem proving by analogy is a process in which the experience of proving a source theorem guides the search for a proof of a similar target theorem. There are at least two different paradigms for analogy: transformational analogy and derivational analogy [2]. Transformational analogy maps a final source solution to the target and may involve additional transformations so the constraints of the target problem are satisfied. Derivational analogy instead maps the problem solving decisions made in the source case [3] and uses these mapped decisions as a ....

J.G. Carbonell. Learning by analogy: Formulating and generalizing plans from past experience. In R.S. Michalsky, J.G. Carbonell, and T.M. Mitchell, editors, Machine Learning: An Artificial Intelligence Approach, pages 371--392. TiogaPubl., Palo Alto, 1983.

....with a target theorem C Th t , where Th s and Th t match. condt introduces C as a subgoal and closes a certain plan branch, if C is disproved. The last kind of reformulations correspond to Carbonell s T(ransformation) operators final segment concatenation and initial segment concatenation [4], where a final plan segment reduces the theorem to a (set of) subgoals and an initial segment transforms assumptions to other proof assumptions. Consider, for instance, the source theorem Th s : 8x:x 2 setA x 2 setB and the target theorem Th t : setA 0 setB 0 . Initially, ....

J.G. Carbonell. Learning by analogy: Formulating and generalizing plans from past experience. In R.S. Michalsky, J.G. Carbonell, and T.M. Mitchell, editors, Machine Learning: An Artificial Intelligence Approach, pages 137--162. Tioga, Palo Alto, 1983.

....knowledge (BK) represented using combinations of XPs and scripts, to explain the story and to build a representation for it in its foreground knowledge (FK) When this task fails, a trace of the reasoning that preceded the failure is passed to the learning subsystem. A case based reasoning [5,34,45,83,87] subsystem within the learner uses past cases of introspective reasoning from the BK to explain the comprehension failure and to generate a set of learning goals. These goals, along with the trace, are then passed to a non linear planner [84,96,100] The planner subsequently builds a learning ....

J. G. Carbonell, Learning by analogy: Formulating and generalizing plans from past experience, in: R. S. Michalski, J. G. Carbonell and T. M. Mitchell, eds., Machine learning I: An artificial intelligence approach (Morgan Kaufmann, Los Altos, CA, 1983) 137-161.

....target inference rules. The problem with this approach is that it finds analogies on a purely syntactic level. This, however, does not suffice for some problems [Owen, 1986] Carbonell has coined the notion of transformational analogy which finds the analogy between two problem representations [Carbonell, 1983]. This is similar to the use of analogy by Munyer. Later, he introduced the notion of derivational analogy [Carbonell, 1986] which instead of looking for analogies between the problem representations alone, also looks for analogies between problem solving activities. The idea is that the entire ....

Carbonell, J.G. (1983). Learning by analogy: Formulating and generalizing plans from past experience. In Michalski, R.S., Carbonell, J.G. and Mitchell, T.M., (eds.), Machine Learning. Tioga Press.

....of primary reaction class, sensitivity analysis, frequency and validation of learning of combination rules) 3.3 Learning Action Rules There are two possible ways to reuse action planning knowledge for future tasks. First, try to transform the result of a planning task according to new needs [5]. Second, build up a set of action rules which determine an appropriate action with respect to the current context. Though the first way should later guide the planning phase by leading to some skeletal goals, our approach now is mainly based on the second method. This is due to the higher ....

J.G. Carbonell. Learning by analogy: Formulating and generalizing plans from past experience. In R.S. Michalski et al., editors, Machine Learning: An Artificial Intelligence Approach Vol. 1, pp. 137--161. Tioga Publishing Company, 1983.

....were studied in the context of machine learning (Bolc, 1987; Michalski et al. 1983; Michalski et.al. 1986) The field of machine learning emphasizes mainly inductive arguments (Michalski, 1983) and eventually combinations of inductive with deductive argument to provide learning by analogy (Carbonell, 1983). Michalski (1987) introduced the following classification of different types of learning: direct learning , learning by instruction, deductive learning, learning by analogy, learning trough examples and learning trough observation and discovery. In the past, machine learning was heavily supported ....

- "Learning by Analogy: Formulating and Generalizing Plans from Past Experience" - in Michalski et.al. Machine Learning - An Artificial Intelligence Approach - Morgan Kaufmann Publishers, Inc. pp 83-134.

....to ensure that no impossibilities are introduced to explanation during modification. Domain independence characterizes the dependence of the system on a particular domain. ffl The system uses only domain dependent rules for adaptation (Kass, 1989; Goel and Chandrasekaran, 1989; Alterman, 1988; Carbonell, 1986). ffl The initial set of techniques is domain independent; later, some domain specific rules are incorporated through the explanation of expectation failures (Converse, Hammond and Marks, 1989) ffl Modification techniques are domain independent (Hinrichs, 1989; Hanks and Weld, 1992) ffl The ....

Carbonell, J. G. (1986). Learning by analogy: Formulating and generalizing plans from past experience. In Michalski, R. S., Carbonell, J. G., and Mitchell, T. M., editors, Machine Learning: An Artificial Intelligence Approach, volume 2.

....Analogy is a heuristic problem solving strategy that guides the problem solving of a target problem, which is similar to a source problem, by using the source problem solving. There are at least two different paradigms for analogy: derivational analogy [Carbonell 86] and transformational analogy [Carbonell 83] Derivational analogy guides the target solution by replaying decisions of the source problem solving process, and it uses information available during this process only, e.g. the justifications for the decisions made. Transformational analogy transforms only the final source solution to ....

....case of xf 1 (f 2 (x) 7 wg(w) which is, however, not a second order substitution. 11 A normalising reformulation introduces a method or subplan on top of the source proof plan. Normalising reformulations resemble the initial segment concatenations in Carbonell s transformational analogy [Carbonell 83] Analogy in CL A M 13 8x 1 8x 2 (x 1 = x 2 f(x 1 ) f(x 2 ) where f is a function or relation. The method reduce could eventually have other parameters (or submethods) that guide its actual work. reduce can be applied if the source theorem is (8)t 1 = t 2 and the target theorem is ....

J.G. Carbonell. Learning by analogy: Formulating and generalizing plans from past experience. In R.S. Michalsky, J.G. Carbonell, and T.M. Mitchell, editors, Machine Learning: An Artificial Intelligence Approach, pages 371--392. TiogaPubl., Palo Alto, 1983.

....Minton s [1988] strategy learning system, which collects statistics to estimate expected utilities. Explicit rationality of learning is also reflected in the goal dependent preference order on generalizations employed by [Stepp and Michalski 1986] in the similarity order on analogies employed by [Carbonell 1983, 1986] and in the preferences guiding shift of bias employed by [Russell and Grosof 1987] See [Doyle 1988c] for more on the relation between rationality, similarity, and shift of bias. 5 Mechanizing rational learning Since decision theory is a general theory intended to cover all sorts of ....

Carbonell, J. G., 1983. Learning by analogy: formulating and generalizing plans from past experience, Machine Learning: An Artificial Intelligence Approach (R. S. Michalski, J. G. Carbonell, and T. M. Mitchell, eds.), Palo Alto: Tioga, 137-161.

.... is a necessary prerequisite for transfer (Novick Holyoak, 1991) For domains such as algebra word problems (Reed, Ackinclose, Voss, 1990; Novick Holyoak, 1991) constructing proofs or programs (Anderson Thompson, 1989) or generating plans (operation sequences to fulfill a given goal; Carbonell, 1983), analogical mapping and transfer are typically described in the following way: mapping is performed on the initial problem descriptions represented as schemes or graphs. Transfer (adaptation) is performed by re instantiation of the source solution with the corresponding concepts in the target ....

Carbonell, J. G. (1983). Learning by analogy: Formulating and generalizing plans from past experience. In R. S.

No context found.

Jaime G. Carbonell. Learning by analogy: Formulating and generalizing plans from past experience. In Ryszard S. Michalski, Jaime G. Carbonell, and Tom M. Mitchell, editors, Machine Learning: An Artificial Intelligence Approach, pages 371--392. Tioga Publishers, Palo Alto, ca, 1983.

No context found.

Machine Learning: An Artificial Intelligence Approach. Tioga Press, Palo Alto, California, 1982.

No context found.

Carbonell J.G.: Learning by Analogy: Formulating and Generalizing Plans from Past Experience. In Machine Learning: An Artificial Intelligence Approach. (1983), 137-161.

No context found.

Carbonell J.G., 1984, "Learning by Analogy: Formulating and Generalizing Plans from Past Experience," in R.S. Michalski, J.G. Carbonell, and T.M. Mitchell, Eds., Machine Learning -- An Artificial Intelligence Approach, Spring-Verlag, Berlin.

No context found.

Carbonell JG. Learning by analogy: formulating and generalizing plans from experience. In: Michalski RS Carbonell JG, Mitchell TM (eds) Machine Learning, vol 1, Tiogo Publishing Company Palo Alto, 1983

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