Using k-d trees to improve the retrieval step in case-based reasoning

Wess, Stefan; Althoff, Klaus-Dieter; Derwand, Guido

doi:10.1007/3-540-58330-0_85

Cited by 105 publications

(51 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This reduces the cost of the search for similar cases (for retrieval or previous to the introduction of new cases in the case-base) to a specific set of cases with the same index as the current case (Patterson et al, 2005) (Quinlin, 1993) (Wess et al, 1993).…”

Section: Temporal Bounded Case-based Reasoningmentioning

confidence: 99%

Towards real-time agreements

Navarro

Heras

Botti

et al. 2013

Expert Systems with Applications

View full text Add to dashboard Cite

In this paper, we deal with the problem of real-time coordination with the more general approach of reaching real-time agreements in MAS. Concretely, this work proposes a real-time argumentation framework in an attempt to provide agents with the ability of engaging in argumentative dialogues and come with a solution for their underlying agreement process within a bounded period of time. The framework has been implemented and evaluated in the domain of a customer support application. Concretely, we consider a society of agents that act on behalf of a group of technicians that must solve problems in a Technology Management Centre (TMC) within a bounded time. This centre controls every process implicated in the provision of technological and customer support services to private or public organisations by means of a call centre. The contract signed between the TCM and the customer establishes penalties if the specified time is exceeded.Keywords: Agreement Technologies, Real-Time, Multi-Agent Systems MotivationMany Multi-Agent Systems (MAS) operate in resource and time-constrained environments. In such domains, scarce resources must be shared between di↵erent agents taking into account that these agents must perform their tasks before a deadline is met. In addition, agents in open MAS are autonomous entities that can have their own knowledge resources, can play di↵erent roles and can have di↵erent objectives and preferences over values that they want to promote with their actions (e.g. an interested agent could want In this paper, we deal with the problem of real-time coordination with the more general approach of reaching real-time agreements in MAS. Therefore, our notion of a real-time agreement may refer to the fair allocation of a scarce resource, but also to the outcome of a negotiation process to make a deal in a sale operation, to the final label assigned in a classification problem, to the final state of the beliefs database of an agent in a beliefs revision problem, and many more. However, any of these processes must be temporal bounded. This is an important aspect to consider in any real-time argumentative process, where a community of agents are able to reach agreements through argumentation before a deadline is met. Any agreement reached after this deadline will be considered of a low quality, or even not valid if we are in a hard real-time environment. Therefore, it is necessary to control every step of the argumentative process in order to predict its duration in time (i.e. we must know the temporal cost of executing each step of the process). Taking into account this prediction, we can guarantee that the agents engaged in the argumentation dialogue will reach an agreement about the outcome of the agreement process (e.g. they will decide the final allocation for a resource, the label of an individual in a classification problem, etc.) before the specified deadline is met. The response provided by the agents does not need to be optimal, since in some cases the best response may require more proce...

show abstract

Section: Temporal Bounded Case-based Reasoningmentioning

confidence: 99%

Towards real-time agreements

Navarro

Heras

Botti

et al. 2013

Expert Systems with Applications

View full text Add to dashboard Cite

show abstract

“…The features are apriori assigned to the dimensions of the hypercube, and the respective values are assigned numeric coordinates within these dimensions. As such, a data object described by a vector <f 1 33 >. To map the description to one of the MLH nodes, the features are hashed to the dimensions of the relevant hypercube, and the values are hashed to the numeric coordinates within these dimensions.…”

Section: Unspecified Data Managementmentioning

confidence: 99%

“…Given a case base containing descriptions of N cases, the number of case distance computations needed for a naïve exhaustive retrieval of most similar cases is O(N). More efficient multidimensional retrieval techniques, such as Kd trees, were proposed in [33]. These technique requires O(N logN) to build the treebased data structure and O(logN) to retrieve the most-similar case in the hypothesis that the number of cases is large, i.e., N>>2 d .…”

Section: Similarity-based Retrieval Over Unsomentioning

confidence: 99%

“…For example, [33] proposes a sound and complete approach to the similarity-based retrieval, in which the organization of the cases is based on the similarities between them. K-d trees [11] are used to split the case base into groups of cases, such that each group contains cases that are similar to each other according to a given similarity measure.…”

Section: Speed-up Techniques and Distributed Cbrmentioning

confidence: 99%

See 1 more Smart Citation

P2P case storage and retrieval with an unspecified ontology

2007

View full text Add to dashboard Cite

Abstract. Traditional similarity-based retrieval of structured data, as in CaseBased Reasoning (CBR) approaches, has been largely implemented using centralized storage systems. In such systems, when the cases or records contain both numeric and symbolic attributes, similarity-based retrieval cannot exploit standard speedup techniques based on multi-dimensional indexing, and the retrieval is implemented by an exhaustive comparison of the case to be solved with the whole set of stored cases. In this work, to improve the performance of the case retrieval step and build CBR systems that can scale up to large case bases, we propose a novel approach for storage of the case base in a decentralized Peer-to-Peer environment using the notion of Unspecified Ontology. We also develop an algorithm for efficient retrieval of approximated most-similar cases, that exploits inherent characteristics of the unspecified ontology in order to improve the performance of the case retrieval step. The experiments show that the algorithm successfully retrieves cases that are very close to the mostsimilar cases, while reducing the number of cases to be compared. Hence, it improves the performance of the retrieval step, the first stage of the CBR problem solving cycle. Moreover, the distributed nature of our approach eliminates the need for a centralized server that not only becomes a computational bottleneck, but is also a single point of failure. Abstract. Traditional similarity-based retrieval of structured data, as in CaseBased Reasoning (CBR) approaches, has been largely implemented using centralized storage systems. In such systems, when the cases or records contain both numeric and symbolic attributes, similarity-based retrieval cannot exploit standard speedup techniques based on multi-dimensional indexing, and the retrieval is implemented by an exhaustive comparison of the case to be solved with the whole set of stored cases. In this work, to improve the performance of the case retrieval step and build CBR systems that can scale up to large case bases, we propose a novel approach for storage of the case base in a decentralized Peer-to-Peer environment using the notion of Unspecified Ontology. We also develop an algorithm for efficient retrieval of approximated most-similar cases, that exploits inherent characteristics of the unspecified ontology in order to improve the performance of the case retrieval step. The experiments show that the algorithm successfully retrieves cases that are very close to the mostsimilar cases, while reducing the number of cases to be compared. Hence, it improves the performance of the retrieval step, the first stage of the CBR problem solving cycle. Moreover, the distributed nature of our approach eliminates the need for a centralized server that not only becomes a computational bottleneck, but is also a single point of failure. Manuscript

show abstract

“…Given a case base containing descriptions of N cases, the number of case dis- More efficient multidimensional retrieval techniques, such as those based on K-d trees [4], were proposed in [15]. K-d tree uses a multi-dimensional tree for management and retrieval of cases.…”

Section: Fig 1 Generalization Of the Fixed Ontology To The Unspecifmentioning

confidence: 99%

P2P Case Retrieval with an Unspecified Ontology

Berkovsky

Kuflik

Ricci³

2005

Case-Based Reasoning Research and Development

View full text Add to dashboard Cite

Abstract. Traditional CBR approaches imply centralized storage of the case base and, most of them, the retrieval of similar cases by an exhaustive comparison of the case to be solved with the whole set of cases. In this work we propose a novel approach for storage of the case base in a decentralized Peer-toPeer environment using the notion of Unspecified Ontology. In our approach the cases are stored in a number of network nodes that is comparable with the number of cases. We also develop an approximated algorithm for efficient retrieval of most-similar cases. The experiments show that the approximated algorithm successfully retrieves the most-similar cases while reducing the number of cases to be compared.

show abstract

Using k-d trees to improve the retrieval step in case-based reasoning

Cited by 105 publications

References 10 publications

Towards real-time agreements

Towards real-time agreements

P2P case storage and retrieval with an unspecified ontology

P2P Case Retrieval with an Unspecified Ontology

Contact Info

Product

Resources

About