Manuela M. Veloso and Jaime G. Carbonell. Derivational analogy in PRODIGY: Automating case acquisition, storage, and utilization. Machine Learning, 10(3):249--278, March 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....knowledge gaps are identified, the reasons why particular hypotheses are generated, the strategies chosen for verifying candidate hypotheses, and the basis for choosing particular reasoning methods for each of these. Trace Meta XPs are similar to reasoning traces [Carbonell, 1986; Minton, 1988; Veloso Carbonell, 1993] or justification structures [Collins, Birnbaum, Krulwich, Freed , 1993; deKleer, Doyle, Steele, Sussman , 1977; Doyle, 1979] with the difference that Trace Meta XPs represent, in addition to the subgoal structure of the problem and justifications for operator selection decisions, ....

M.M. Veloso and J.G. Carbonell. Derivational Analogy in PRODIGY: Automating Case Acquisition, Storage, And Utilization. Machine Learning, 10(3):249--278, 1993.

....is not new to the field of robotics. It was successfully used to help in solving such problems as path planning based on past routes, high level action selection based on environment similarities, place learning, and acceleration of complex problem solving based on past problem solutions [3, 4, 5, 6, 7, 8]. Previous work has also been performed on the incorporation of case based reasoning in the selection of behavior parameters by our group [1, 2] on which this present research is partially based and a few others [e.g. 13] The approach described in this paper, however, differs significantly from ....

M.M. Veloso and J. G. Carbonell, "Derivational Analogy in PRODIGY: Automating Case Acquisition, Storage, and Utilization," Machine Learning, 10(3), pp. 249-278, 1993.

....is not new to the field of robotics. It was successfully used to help in solving such problems as path planning based on past routes, high level action selection based on environment similarities, place learning, and acceleration of complex problem solving based on past problem solutions [4, 5, 6, 7, 8, 14]. Previous work has also been performed on the incorporation of case based reasoning in the selection of behavior parameters by our group [1, 2, 3] upon which this present research is partially based) as well as a few other groups [e.g. 13] The approach described in this paper extends our ....

M.M. Veloso and J. G. Carbonell, "Derivational Analogy in PRODIGY: Automating Case Acquisition, Storage, and Utilization," Machine Learning, 10(3), pp. 249-278, 1993.

....directly. Thus, transformational planners and schedulers are more flexible, but this comes at a price: They are also more complicated algorithms and can have worse performance if not carefully controlled. Another use for transformational planners is in case based planning [Hammond 1989, Veloso and Carbonell 1993]. In case based planning, a library of previously used plans is maintained. When a new set of goals is presented, the algorithm searches the library for a similar problem that it has solved in the past. The plan retrieved is then modified, using transformational methods, to achieve the new set of ....

M. Veloso and J. Carbonell. Derivational Analogy in PRODIGY: Automating Case Acquisition. Machine Learning, Vol. 10, pages 249-278. 1993.

....the size of the search space generated by a planning system. 1 Introduction As reported in several independent experiments, case based planners using Derivational Analogy have consistently outperformed the base level, first principles planner on which these case based planners were constructed [1 3]. On the other hand, formal studies on the complexity of adaptation versus the complexity of first principles planning seem to suggest that in the worst case, adaptation can be harder than planning from scratch if certain conditions on the adaptation strategy are satisfied [4] These complexity ....

....#p t2# #q t2# Precondition Add list Delete List o # p, q G # o i i # i1 oq q p Fig. 1. An example of plan generation with UCP in [5] 3 3 DerUCP: Derivational Analogy in UCP Derivational Analogy is a widely used adaptation method that has been the subject of frequent studies [1, 2, 7 12]. In Derivational Analogy cases contain the derivational trace, the sequence of decisions made to obtain a plan, rather than the plan itself. Typically in a problem solving session, part of a solution plan is obtained through case replay of the derivational traces stored in retrieved cases, and ....

[Article contains additional citation context not shown here]

Veloso, M., Carbonell, J.: Derivational analogy in PRODIGY: Automating case acquisition, Storage and Utilization. Machine Learning (1993) 249--278

....in isolation, but the interplay between these methods is also being investigated. This paper focuses on the interaction between two such learning methods: case based reasoning (CBR) and learning momentum (LM) Both methodologies were successfully used in robotic systems in different contexts [2, 3, 8, 9, 10, 11, 12, 13, 14]. In this work these methods are used to change behavioral parameters of a behaviorbased robotic system at run time. Both algorithms have already been shown, in isolation, to increase performance in a robotic system in relation to navigating unknown obstacle fields while trying to reach a goal ....

Veloso, M. M., Carbonell, J. G., "Derivational Analogy in PRODIGY: Automating Case Acquisition, Storage, and Utilization," Machine Learning, Vol. 10, No. 3, 1993, pp. 249-278.

....is not new to the field of robotics. It was successfully used to help in solving such problems as path planning based on past routes, high level action selection based on environment similarities, place learning, and acceleration of complex problem solving based on past problem solutions [4, 5, 6, 7, 8, 14]. Previous work has also been performed on the incorporation of case based reasoning in the selection of behavior parameters by our group [1, 2, 3] upon which this present research is partially based) as well as a few other groups [e.g. 13] The approach described in this paper extends our ....

M.M. Veloso and J. G. Carbonell, "Derivational Analogy in PRODIGY: Automating Case Acquisition, Storage, and Utilization," Machine Learning, 10(3), pp. 249-278, 1993.

....plans with respect to different quality measures. One way to improve planning in a given domain is to learn search control knowledge to guide the planning process. There have been different approaches to acquiring control knowledge for non trivial (nonlinear) planning. Some of them use analogy [7, 17], others deduction [8, 12] induction [10] and some combine deduction and induction [2, 5] Fully deductive methods have their weaknesses. For instance, they usually require a complete domain theory. Alternatively, inductive approaches incrementally acquire correct knowledge by observing a large ....

Manuela M. Veloso and Jaime G. Carbonell. Derivational analogy in PRODIGY: Automating case acquisition, storage, and utilization. Machine Learning, 10(3):249--278, March 1993.

....solved by decreasing plan quality, but this is not necessarily so. Also, the Hamlet seed bias Hamlet EvoCK conveniently with respect to plan quality. 5 Related Work There have been di erent approaches to acquire control knowledge for nontrivial (non linear) planning. Some of them use analogy [15,42], others pure deduction (EBL) 16,17,26] pure induction [23] and some combine deduction (EBL) and induction like Hamlet [5] and SCOPE [8,9] EBL ILP) However, they do not use genetic search as a component for the multi strategy system. In particular, 8] o ers many results that could be used ....

Manuela M. Veloso and Jaime G. Carbonell. Derivational analogy in PRODIGY: Automating case acquisition, storage, and utilization. Machine Learning, 10(3):249-278, March 1993.

....machine learning, etc. In order to explore eciently such (usually huge) state spaces, di erent Machine Learning (ML) techniques have been devised. Some of them learn global Preprint submitted to Elsevier Preprint 27 March 2001 heuristics such as macro operators [10,19,32] chunks [37] or cases [13,14,41]. Other techniques learn local search heuristics (or control knowledge) 8,26] All machine learning algorithms have biases that determine how they generalize to unseen instances [38] Sometimes it is desirable that they possess other learning biases. However, it might be dicult to add those ....

Manuela M. Veloso and Jaime G. Carbonell. Derivational analogy in prodigy: Automating case acquisition, storage, and utilization. Machine Learning, 10:249-278, 1993.

....feedback. CELIA initially selects cases using a wide range of possible features #Redmond, 1992#. It prunes the features for a case when feedback from a human expert indicates the case is Background 32 inapplicable. It also alters the indices to retrieve a case whichwas incorrectly omitted. Veloso Carbonell #1993# use derivational analogy to produce solutions by examining the stored reasoning traces of previous solutions, in PRODIGY. PRODIGY s memory manager suggests cases to its problem solving component, and alters the indices for the retrieved cases based on positive or negative feedback from the ....

Veloso, M. & Carbonell, J. #1993#. Derivational analogy in prodigy: automating case acquisition, storage, and utilization. Machine Learning, 10 #3#, 249#278.

....frameworks. However, other areas of machine learning have seen a few frameworks of this sort. Langley and Neches [7] developed Prism, a flexible language for production system architectures that supported many combinations of performance and learning algorithms, and later versions of Prodigy [14] included a variety of mechanisms for learning search control knowledge. For classification problems, Kohavi et al. s [5] MLC supported a broad set of supervised induction algorithms that one could invoke with considerable flexibility. The generative abilities of MLC are apparent from its use ....

M. M. Veloso and J. G. Carbonell. Derivational analogy in Prodigy: Automating case acquisition, storage, and utilization. Machine Learning, 10:249--278, 1993.

....di erent PlannerAgents [5] would allow to distribute the computational cost of the planning process. In the domain considered in this paper, it seems easy to develop distributed approaches because the problems considered have independent goals. 3. To develop Case Based Planning Skills (CbP) [11]. The PlannerAgents would store old successful plans and use Case Based Reasoning techniques, so that new planning problems can be solved by adapting previously solved plans which were similar. This would reduce enormously the planning process which might become computationally expensive. Even if ....

Veloso, M., Carbonell, J.: Derivational Analogy in PRODIGY: Automating Case Acquisition, Storage, and Utilization. Machine Learning (1993). Volume 10 (3), 249{ 278.

No context found.

Manuela M. Veloso and Jaime G. Carbonell. Derivational analogy in PRODIGY: Automating case acquisition, storage, and utilization. Machine Learning, 10(3):249--278, March 1993.

No context found.

Veloso, M. M., and Carbonell, J. G. 1993. Derivational analogy in PRODIGY: Automating case acquisition, storage, and utilization. Machine Learning 10:249--278.

No context found.

Manuela M. Veloso and Jaime G. Carbonell. Derivational analogy in PRODIGY: Automating case acquisition, storage, and utilization. Machine Learning, 10(3):249--278, March 1993.

....which the rules are to be transferred to individual decision steps in other problems. Alternative learning approaches in nonlinear planning include learning complete generalized plans as in [12] or developing a case based learning method that provides cases as a form of global strategic knowledge [24], as discussed in the related work section. 2.2 hamlet s Components hamlet has three main modules: the Bounded Explanation learner, the Inducer and the Refiner. The Bounded Explanation module learns control rules from the search tree. These rules are either over specific or over general, so they ....

Manuela M. Veloso and Jaime G. Carbonell. Derivational analogy in prodigy: Automating case acquisition, storage, and utilization. Machine Learning, 10:249-- 278, 1993.

....this process has to be repeated for each metric. We believe that a better solution consists on automatically learning those heuristics by experience on solving previous problems. For this purpose, one could try to use any of the previous work on search control knowledge. Some of them use analogy [10,19], others pure deduction [11,13] pure induction [12] and some combine deduction and induction [5,7] However, while planning to obtain a solution plan has been largely studied in the literature, the search for optimal or good plans has been usually discarded due to its complexity. During the last ....

Manuela M. Veloso and Jaime G. Carbonell. Derivational analogy in PRODIGY: Automating case acquisition, storage, and utilization. Machine Learning, 10(3):249-278, March 1993.

No context found.

Veloso, M. M., and Carbonell, J. G. Derivational analogy in PRODIGY: Automating case acquisition, storage, and utilization. Machine Learning, 10, 249--278, 1993.

No context found.

Manuela M. Veloso and Jaime G. Carbonell. Derivational analogy in prodigy: Automating case acquisition, storage, and utilization. Machine Learning, 10:249--278, 1993. 42

No context found.

Manuela M. Veloso and Jaime G. Carbonell. Derivational analogy in PRODIGY: Automating case acquisition, storage, 19 and utilization. Machine Learning, 10(3):249-278, March 1993.

No context found.

Veloso, M. and J. Carbonell: 1993, `Derivational Analogy in PRODIGY: Automating Case acquisition, storage, and utilization'. Machine Learning 10, 249--278.

No context found.

M. Veloso and J. Carbonell. Derivational analogy in prodigy: Automating case acquisition, storage and utilization. In Machine Learning, pages 249--278, 1993. 13

No context found.

Veloso, M. M. & Carbonell, J. G.(1993). Derivational Analogy in PRODIGY: Automating Case Acquisition, Storage, and Utilization. Machine Learning, 10(3):249-278.

No context found.

Manuela Veloso and Jaime Carbonell. Derivational analogy in PRODIGY: Automating case acquisition, storage, and utilization. Machine Learning, 10:249--278, 1993.

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