Xuemei Wang. Learning planning operators by observation and practice. In Proceedings of the Second International Conference on AI Planning Systems, AIPS-94, pages 335--340, Chicago, IL, June 1994. Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
First 50 documents
Learning Domain Knowledge for Teaching Procedural Skills - Angros, Jr., Johnson.. (2002) (1 citation) (Correct)
....DEMONSTRATION, EXPERIMENTATION, AND DIRECT SPECIFICATION Diligent builds on ideas from a variety of prior systems, but these prior systems have placed a greater burden on instructors. Some ask a large number of questions [31, 29, 10, 24] some require a large number of demonstrations [32, 33, 17], and some require an initial domain theory [6, 7, 20, 13, 25, 14, 19] Diligent minimizes the need for demonstrations, questions, and an initial domain theory through a novel combination of programming by demonstration, autonomous experimentation, and direct specification. 3.1 Demonstrations ....
X. Wang. Learning Planning Operators by Observation and Practice. PhD thesis, Carnegie Mellon University, 1996.
Learning Domain Knowledge for Teaching Procedural Tasks - Scholer, Rickel, Jr., Johnson (2000) (1 citation) (Correct)
....an intelligent agent. Systems have been designed that use higher level instruction languages to make knowledge engineering quicker and more easily understood (Badler, 1998) Alternatively, some systems seek to eliminate traditional programming through instruction (Huffman 1994) example solutions (Wang 1996) or experimentation (Gil 1992) Most of this work focuses on providing agents with the knowledge to perform tasks as opposed to teaching the tasks. While learning how to do and how to teach how to do are similar problems, an agent that is to be an instructor has an extra requirement it must be ....
Wang , X. 1996. Learning Planning Operators by Observation and Practice. Ph.D. diss., School of Computer Science, Carnegie Mellon Univ.
Learning Hierarchical Performance Knowledge by Observation - van Lent, Laird (1999) (1 citation) (Correct)
....Thus the time steps can potentially be very short with lots of activity in some time steps and no activity in others. In the next section two supervised learning systems which use observation as the primary knowledge source are described, behavioral cloning [1, 6] and the OBSERVER system [8, 9]. Section three describes the format of the performance knowledge generated. The KnoMic system, described in the fourth section, combines many of the strengths of the systems described in section two while overcoming some of their weaknesses. The fifth section provides examples and evaluation of ....
Wang, X. (1996). Learning Planning Operators by Observation and Practice. PhD thesis, Carnegie Mellon University, Computer Science Dept., 1996.
Recovering from Modeling Faults in GOLOG - Bjäreland (Correct)
.... as Given a detected discrepancy, when is it possible to correctly classify the discrepancy as being due to a faulty model (faulty expectation) This question has also, though to a smaller degree, interested researchers concerned with machine learning of planning operators (e.g. Benson, 1996; Wang, 1994 ] Their work has focused on how to improve the models with learning techniques from discrepancies (negative instances) and positive instances provided by a teacher. The classification in theses approaches are typically performed when the same operator has failed repeatedly, e.g. in Benson s ....
X. Wang. Learning planning operators by observation and practice. In AIPS94.
Planning under Uncertainty by Spreading Activation through an.. - Bagchi (1994) (Correct)
....food using a fork increases when the plate is full, whereas the probability of using a spoon is larger when the plate is close to empty. Typically, inductive learning mechanisms require a large number of examples, where an action is performed under a variety of situations (Vere 1978; Kadie 1988; Wang 1994). This is to ensure that coincidental associations do not produce spurious high probability values. Introduction 3 This research extends previous work by analyzing how erroneous domain knowledge learned may be learned when an action is always observed in the context of the same plan, and then ....
....be established categorically because of causes (1) and (2) above. Instead, probabilistic relations based on correlations have to be established. Our objective of learning action preconditions and effects from observation is similar to those in the past (Vere 1978; Kadie 1988; Carbonell Gil 1990; Wang 1994) (see Chapter 2) However our approach of using correlational measures to obtain a probabilistic relation between actions and propositions is different from the past approaches, which make the assumption of deterministic domains. In terms of the domain knowledge representation described in Chapter ....
[Article contains additional citation context not shown here]
Wang, X. Learning planning operators by observation and practice. In Proceedings of the Second International AI Planning Conference. Chicago, IL: AAAI Press, 1994.
Mining GPS Data to Augment Road Models - Rogers, Langley, Wilson (1999) (1 citation) (Correct)
....in the world, then analyze the effects to determine how to perform better in the future. Our problem is fundamentally different because our system cannot to perform experiments to test hypotheses. Instead, it is forced to passively observe the environment and build knowledge structures. Although Wang s (1996) OBSERVER also passively observes a series of expert execution traces, it also requires sensors to record the effects of the expert s actions on goal conditions. A final distinction from all these planning systems is that, rather than accomplish any specific goal, our system attempts to augment ....
Wang, X. (1996). Learning planning operators by observation and practice. Doctoral dissertation, School of Computer Science, Carnegie Mellon University, Pittsburgh, PA.
Learning Procedural Planning Knowledge In Complex Environments - Pearson (1996) (8 citations) (Correct)
....Extensional Semi Extensional IMPROV Intentional Extensional Semi Extensional Declarative O(size of rep) Procedural O(1) Theory Revision Systems Figure 3. 3: Classification of learning systems EXPO, LIVE and OBSERVER EXPO [Gil, 1991; Gil, 1993] LIVE [Shen and Simon, 1989] and OBSERVER [Wang, 1995; Wang, 1996] share a similar STRIPS like [Fikes and Nilsson, 1971] representation. Operators are declaratively represented as structures with lists of preconditions and effects. The preconditions are represented in disjunctive normal form. The effects are limited to a single, state to state transition, ....
....to other operators and generalization of previous experiences combined with a module to efficiently design experiments. Once the cause of the failure has been determined, EXPO changes the operator preconditions accordingly. LIVE [Shen, 1989; Shen, 1994] and the more recent OBSERVER [Wang, 1995; Wang, 1996] learn domain knowledge by executing actions in the environment and observing the results. LIVE and OBSERVER rely on the assumption that changes in the environment are due to deterministic actions of the agent. Differences between the state before and then after an action are used as the basis ....
[Article contains additional citation context not shown here]
Xuemei Wang. Learning Planning Operators by Observation and Practice. PhD thesis, Carnegie Mellon University, 1996.
Learning to Predict Lane Occupancy Using GPS and Digital Maps - Rogers, Langley, Wilson (Correct)
....in the world, then analyze the effects to determine how to perform better in the future. Our problem is fundamentally different because our system cannot to perform experiments to test hypotheses. Instead, it is forced to passively observe the environment and build knowledge structures. Although Wang s (1996) OBSERVER also passively observes a series of expert execution traces, it also requires sensors to record the effects of the expert s actions on goal conditions. A final distinction from all these planning systems is that, rather than accomplish any specific goal, our system attempts to augment ....
Wang, X. (1996). Learning planning operators by observation and practice. Doctoral dissertation, School of Computer Science, Carnegie Mellon University, Pittsburgh, PA.
Learning by Observation and Practice: Towards Real.. - Wang, Carbonell (1994) (1 citation) Self-citation (Wang) (Correct)
....number of practice problems Prodigy with correct operators learning system Figure 5: Summary In summary, learning, planning, plan repair, and plan execution are tightly integrated in our learning system, and operators are learned through a specific to general inductive generalization process. (Wang 1994) provides more details and examples of the learning algorithm. The empirical results demonstrate the effectiveness of our approch. 0 1000 2000 3000 4000 5000 6000 7000 8000 number of nodes 0 5 10 15 20 25 number of practice problems Prodigy with correct operators learning system Figure 6: On ....
Wang, X. 1994. Learning Planning Operators by Observation and Practice. In Proceedings of the Second International Conference on AI Planning Systems.
A Multistrategy Learning System for Planning Operator Acquisition - Wang (1996) (1 citation) Self-citation (Wang) (Correct)
....The planner also repairs the failed plans upon unsuccessful executions. The repaired plans are then executed in the environment. This process repeats until the problem is solved, or until a resource bound is exceeded. Details of the integration of planning, execution, and learning can be found in (Wang 1996b) Refining operators during practice: The successful and unsuccessful executions generated during practice are effective training examples that OBSERVER uses to further refine the initial imperfect operators. Details for operator refinement during practice are described in this paper. Issues ....
....effects of op if missing effects are observed in the delta state of the execution. OBSERVER then updates the current state and continues execution with the remaining plan. Detailed descriptions of the algorithms for planning with incomplete and incorrect operators and plan repair are described in (Wang 1996b) Refining operators during practice This section describes how OBSERVER refines operators using the successful and unsuccessful executions generated during practice. This involves updating the S rep of operator preconditions, learning the G rep of the operator preconditions, and updating the ....
[Article contains additional citation context not shown here]
X. Wang. Learning Planning Operators by Observation and Practice. PhD thesis, School of Computer Science, Carnegie Mellon University, Pittsburgh, PA, 1996.
Planning While Learning Operators - Wang (1996) (12 citations) Self-citation (Wang) (Correct)
.... most general boundary (the G set) in version spaces in several ways: i) they are not guaranteed to be the most specific or general representations consistent with the set of positive and negative examples, ii) they are both single boundaries, and (iii) they can be updated in polynomial time (Wang 1996). Domain Knowledge Imperfections Domain knowledge can be incomplete or incorrect in the following ways, as discussed in (Huffman, Pearson, Laird 1993) over general preconditions, over specific preconditions, incomplete effects, extraneous effects, and missing operators. As a result of our ....
Wang, X. 1996. Learning Planning Operators by Observation and Practice. Ph.D. Dissertation, School of Computer Science, Carnegie Mellon University, Pittsburgh, PA.
Learning by Observation and Practice: An Incremental Approach for.. - Wang (1995) (20 citations) Self-citation (Wang) (Correct)
....plans, providing the learning module with the execution traces and passing the plan failures to the planning module for plan repair. This paper concentrates on describing the learning module. Detailed descriptions of the planning module, and the integration of different modules can be found in [Wang, 1994] . When OBSERVER learns deterministic STRIPS like operators, it assumes that the operators have conjunctivepreconditions. This greatly reduces the search space for operator preconditions without sacrificing much of the generality of learning approach, since in most application domains, the ....
....change due to unmet preconditions. The planning module repairs the plan upon execution failures by using S(Op) to determine which additional preconditions to achieve in order to make the failed operator applicable. Detailed descriptions of the planning and plan repair algorithms can be found in [Wang, 1994, Wang and Carbonell, 1994] 4 Learning algorithm descriptions This section presents details of OBSERVER s learning module for learning the preconditions and effects of operators from observation and practice. The observation traces of an expert (denoted as OBS) consists of the pre state and ....
X. Wang. Learning planning operators by observation and practice. In Proceedings of the Second International Conference on AI Planning Systems, Chicago, IL, 1994.
Acquiring Domain-Specific Planners by Example - Elly Winner Manuela (Correct)
No context found.
Xuemei Wang. Learning planning operators by observation and practice. In Proceedings of the Second International Conference on AI Planning Systems, AIPS-94, pages 335--340, Chicago, IL, June 1994.
DISTILL: Towards Learning Domain-Specific Planners by Example - Elly Winner And (Correct)
No context found.
Wang, X. 1994. Learning planning operators by observation and practice. In Proceedings of the Second International Conference on AI Planning Systems, AIPS-94, 335-- 340.
Prodigy Bidirectional Planning - Fink, Blythe (Correct)
No context found.
Xuemei Wang. Learning Planning Operators by Observation and Practice. Report cmu-cs-96-154.
Prodigy Bidirectional Planning - Fink, Blythe (Correct)
No context found.
Xuemei Wang. Learning planning operators by observation and practice. In Proceedings of the Second International Conference on Artificial Intelligence Planning Systems, pages 335--340, 1994.
Learning and Using Models of Kicking Motions for Legged Robots - Sonia Chernova And (Correct)
No context found.
X. Wang, "Learning planning operators by observation and practice," in Proceedings of the Second International Conference on AI Planning Systems, AIPS-94, Chicago, IL, June 1994, pp. 335--340.
Analogical Path Planning - Saul Simhon Gregory (Correct)
No context found.
X. Wang, "Learning planning operators by observation and practice," in Artificial Intelligence Planning Systems, 1994, pp. 335--340.
Path Planning Using Learned Constraints and Preferences - Gregory Dudek And (2003) (Correct)
No context found.
X. Wang, "Learning planning operators by observation and practice," in Artificial Intelligence Planning Systems, pp. 335--340, 1994.
Learning Refinements on Curve-Strokes - Saul Simhon Gregory (Correct)
No context found.
X. Wang. Learning planning operators by observation and practice. In Artificial Intelligence Planning Systems, pages 335--340, 1994.
Some Approaches to Learning in Problem Solving - Millan (Correct)
No context found.
Xuemei Wang. Learning planning operators by observation and practice. In Proceedings of the Second International Conference on AI Planning Systems, AIPS-94, pages 335--340, Chicago, IL, June 1994. AAAI Press, CA.
Analogical Path Planning - Saul Simhon Gregory (Correct)
No context found.
X. Wang, "Learning planning operators by observation and practice," in Artificial Intelligence Planning Systems, 1994, pp. 335--340.
Path Planning Using Learned Constraints and Preferences - Gregory Dudek And (2003) (Correct)
No context found.
X. Wang, "Learning planning operators by observation and practice," in Artificial Intelligence Planning Systems, pp. 335--340, 1994.
Learning Procedural Knowledge through Observation - van Lent, Laird (2001) (4 citations) (Correct)
No context found.
X. Wang. Learning Planning Operators by Observation and Practice. PhD thesis, Computer Science Dept., Carnegie Mellon University, 1996.
Acquiring Domain-Specific Planners by Example - Winner, Veloso (2003) (Correct)
No context found.
Xuemei Wang. Learning planning operators by observation and practice. In Proceedings of the Second International Conference on AI Planning Systems, AIPS-94, pages 335-340, Chicago, IL, June 1994.
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