Buntine, W., "Graphical Models for Discovering Knowledge", in U. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy, editors, Advances in Knowledge Discovery and Data Mining, pp 59-82. AAAI/MIT Press, 1996.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... Markov Chain Monte Carlo Methods, Bayesian smoothing Part of this paper has been presented at the ECAI 96 conference in Budapest 1 Introduction Probabilistic graphical models have been recognized as one of the key paradigms in Intelligent Data Analysis, Data Mining and Knowledge Discovery [1]. In this paper we will devote our attention to a well known class of graphical models, called Causal Prob abilistic Networks (CPNs) or Bayesian Networks [2] These models are Directed Acyclic Graphs in which the nodes represent the (random) variables of the problem, while arcs define ....

Buntine, W. L., Graphical Models for discovering knowledge, in Influence Dia- grams, Belief Nets and Decision Analysis, (U. Fayyad, G. Plateletsky-Shapiro, P. Smyth, and R. Uthurusamy, eds.), AAAI Press, Menlo Park, 1996.

....understandable) reducing run time by solving smaller problems and by using parallel or distributed computation and allowing different solution techniques for individual sub problems. The decomposition approach is frequently used for operations research and engineering design, however, as Buntine [7] states, it has not attracted as much attention in KDD and machine learning community. Most of the work in machine learning decomposition can be found either in practical attempts in specific real life applications (see [7] or in treatments of closely related problems mainly in the context of ....

....used for operations research and engineering design, however, as Buntine [7] states, it has not attracted as much attention in KDD and machine learning community. Most of the work in machine learning decomposition can be found either in practical attempts in specific real life applications (see [7]) or in treatments of closely related problems mainly in the context of distributed and parallel learning (see [42] or multiple classifiers (see [1] Figure 1.0 illustrates our approach for arranging the different types of decomposition in supervised learning. Fig. 1. Decomposition methods in ....

[Article contains additional citation context not shown here]

Buntine, W., "Graphical Models for Discovering Knowledge", in U. Fayyad, G. PiatetskyShapiro, P. Smyth, and R. Uthurusamy, editors, Advances in Knowledge Discovery and Data Mining, pp 59-82. AAAI/MIT Press, 1996.

.... Given independent identically distributed data, we may write the conditional probability of given for a HME model with levels This belief network formalism is based on that of Jordan et al. 110] Other Bayesian or belief network interpretations of mixtures of experts are given by Buntine [28] and Ripley [195] 2.2. HIERARCHICAL MIXTURES OF EXPERTS 14 2 1 Figure 2.3: Belief network for the hierarchical mixture of experts. The target variable is dependent on the input and each of the hidden states . These states replace the single hidden state mixtures of experts. Each ....

Buntine, W. L. [1995], Graphical models for discovering knowledge, in Fayyad, Piatetsky-Shapiro, Smyth and Uthurusamy [52], chapter 3, pp. 59--83. URL: http://www.ultimode.com/wray/ BIBLIOGRAPHY 181

....achieving clearer results, reducing run time by solving smaller problems and by using parallel computation, and allowing different solution techniques for individual sub problems. The decomposition approach is frequently used for operation research and engineering design, however, as Buntine [3] states, it has not attracted as much attention in KDD and machine learning. In this paper we present the attribute decomposition approach where the original attribute set is decomposed to mutually exclusive subsets by an algorithm. A learning algorithm is run upon the training data for each ....

W. Buntine. Graphical Models for Discovering Knowledge, in U. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy, editors, Advances in Knowledge Discovery and Data Mining, pp 59-82. AAAI/MIT Press, 1996.

.... based function [96, 97] the backpropagation algorithm for artificial neural networks (ANNs) 77, 111] decision tree construction algorithms utilizing various node decision criteria [29, 134, 5] spline methods for classification [165, 162] and probabilistic graphical dependency models [76, 32]. Evaluating an estimate g of g in terms of how well it performs on data not included in the training set, or how well g generalizes, is paramount. Often it is possible to allow a learning algorithm to construct g from a sufficiently complex function class so that g approximates g arbitrarily ....

.... to as batch and online k Mean clustering) Kohonen maps [87] and competitive learning [141] Probabilistic clustering methods include the COBWEB algorithm [60] AutoClass [39] the Expectation Maximization (EM) algorithm [50, 124] and, more recently, probabilistic graphical approaches [76, 32]. Hard clustering algorithms, such as k Mean, assign data items to a single cluster whereas soft clustering algorithms, such as EM, assign a given data point to all clusters with a certain probability of membership. Hard clustering algorithms can often be placed in the probabilistic framework ....

W. L. Buntine. Graphical models for discovering knowledge. In Advances in Knowledge Discovery and Data Mining, pages 59--81, Menlo Park, CA, 1996. AAAI Press.

....such as aircraft or trains. Fung and Favero, 1995] describe an Application of Bayesian networks in the retrieval of information, according to the users areas of interest. Burnell and Horvitz, 1995] present a system that utilizes a Bayesian network for debugging very complex computer programs. [Buntine, 1996] approaches the utilization of Bayesian networks in coding, representing and discovering knowledge, through some processes that seek new knowledge on a given domain based on inferences on new data and or on the knowledge already available [Kl osgen, 1996] Hood and Ji, 1997] utilize Bayesian ....

Wray Buntine. Graphical Models for Discovering Knowledge, pages 59--82. In Fayyad et al. [1996], 1996. 134

.... the most strongly associated pairs of events in execution traces; CHAID analysis determines which are most strongly associated [5] A variety of other methods have been developed for generating graphical models, especially Bayesian Networks and Markov models, most of which are for acyclic graphs [4]. Business and medical applications dominate the applications of data mining methods to event sequence modeling. The Quest system at IBM has been used in business for attached mailing, add on sales and customer satisfaction as well as medical diagnosis [2] The algorithm for finding sequential ....

Wray Buntine. Graphical models for discovering knowledge. In U. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy, editors, Advances in Knowledge Discovery and Data Mining. AAAI Press, Menlo Park, CA, 1996.

....655721, fax: 44 (1908) 653169, email: M.Ramoni open.ac.uk, url: http: kmi.open.ac.uk marco. Discovering Bayesian Networks in Incomplete Databases 1. Introduction Bayesian Belief Networks (bbns) are becoming increasingly popular in the Knowledge Discovery and Data Mining (kdd) community [2, 9]. bbns are a successful knowledge representation and reasoning formalism based on probability theory. A bbn [12] is defined by a graphical structure of conditional dependencies among the domain variables and a set of probability distributions defining these dependencies. In this way, bbns provide ....

W. Buntine. Graphical models for discovering knowledge. In Advances in Knowledge Discovery and Data Mining, pages 59--81. MIT Press, Cambridge, MA, 1996. Discovering Bayesian Networks in Incomplete Databases

....will be reported. AUTOCLASS can find hierarchical clusterings and the local maxima are usually sufficient. However, it deals only with simple values. The use of Bayesian networks (Directed Acyclic Graph or DAG) for the discovery of causal relationships among objects is proposed for KDD by Buntine [11, 12], Spirtes et al. 103] and Hackerman et al. 54, 55] Nodes in a Bayesian network represent variables or states, and arcs represent the dependencies between nodes, directed from the cause to the effect. Figure 2.1 gives a very simple Bayesian network for medical problems [11] CHAPTER 2. ....

....KDD by Buntine [11, 12] Spirtes et al. 103] and Hackerman et al. 54, 55] Nodes in a Bayesian network represent variables or states, and arcs represent the dependencies between nodes, directed from the cause to the effect. Figure 2. 1 gives a very simple Bayesian network for medical problems [11]. CHAPTER 2. RELATED WORK IN KDD 18 Age Occupation Climate Symptoms Disease Figure 2.1: A simple Bayesian network. There are usually three steps in constructing a Bayesian network. They involve: ffl Deciding which variables to be modeled as nodes. ffl Determining the structure of the DAG, for ....

W. Buntine. Graphical models for discovering knowledge. In U.M. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy, editors, Advances in Knowledge Discovery and Data Mining, pages 59--82. AAAI/MIT Press, 1996.

....ffl Summarization. Summarization involves techniques for searching compact descriptions for a set of data. Summarization rules, linguistic summaries [123] statistical summaries [52] and decision trees [22] are the main summarization methods used in Data Mining. ffl Dependency modelling [14, 45, 51, 90]. This task looks for a model that describes in an easily understandable way the dependencies that exist between the variables appearing in the data. Two level of analysis are performed in the learning of this kind of models. In a first level we have to find the structural dependence between ....

Wray Buntine. Graphical Models for Discovering Knowledge. In Smyth Fayyad, PiatetskyShapiro and Uthurusamy, editors, Advances in Knowledge Discovery and Data Mining. MIT Press, 1996.

.... independent identically distributed data, we may write the conditional probability of Y given X for a HME model with 3 levels 3 This belief network formalism is based on that of Jordan et al. 110] Other Bayesian or belief network interpretations of mixtures of experts are given by Buntine [28] and Ripley [195] 2.2. HIERARCHICAL MIXTURES OF EXPERTS 14 W 2 W 3 Y X W 1 Figure 2.3: Belief network for the hierarchical mixture of experts. The target variable Y is dependent on the input X and each of the hidden states W 1 , W 2 and W 3 . These states replace the single hidden ....

Buntine, W. L. [1995], Graphical models for discovering knowledge, in Fayyad, Piatetsky-Shapiro, Smyth and Uthurusamy [52], chapter 3, pp. 59--83. *http://www.ultimode.com/wray/ BIBLIOGRAPHY 181

....new techniques for deriving relations between variables are being developed (Djoko, Cook, Holder 1995; Dzeroski 1996) Understandability of patterns: In many applications it is important to make the discoveries more understandable by humans. Possible solutions include graphical representations (Buntine 1996; Heckerman 1996) rule structuring, natural language generation, and techniques for visualization of data and knowledge. Rule refinement strategies (e.g. Major Mangano 1995) can be used to address a related problem: the discovered knowledge may be implicitly or explicitly redundant. User ....

Buntine, W. 1996. Graphical Models for Discovering Knowledge, in AKDDM, AAAI/MIT Press, 59--82.

.... Second, Bayesian networks provide a ready, unifying specification language, as seen by their widespread use in communities such as applied Bayesian statistics and neural information processing [17, 8] their role for the data mining community is to provide a flexible data modeling language [4]. Finally, program synthesis has been proven to be competitive in other domains. It offers: ffl Rapid turn around : even for large tasks mature synthesis systems usually require less than a few minutes to produce code [11, 2] ffl Reliability : synthesized code is used in production systems to ....

....the conditions of Lemma 2; moreover, U 0 is just f cg which has the same dimensions as the given data vector x. This condition triggers the EM algorithm as described in Section 2. 1, and instantiates its schema, resulting in the partial program: while(converging( oe; ae) for( i,j) 1,200] [1,4]) q i;j = P r(c i = jjx i = j; oe) optimize f; oe; aeg for LogP r(x i ; c i j c i ; oe c i ; ae) given fc i q i; xg In this, converging is a generic convergence criterion imposed over the variables ; oe; ae. Given c i q i; implies we quantify c i out of the objective by averaging. ....

[Article contains additional citation context not shown here]

W. Buntine. Graphical models for discovering knowledge. In U. M. Fayyad et al. (eds.), Advances in Knowledge Discovery and Data Mining. MIT Press, 1995.

....scientists, for instance in artificial intelligence, have often contributed more in terms of combining and scaling up these techniques, and generalizing them to classes of representations. More examples of the variety of probabilistic networks and their applications to learning are given in [23] [27]. Learning of probabilistic networks includes a number of complications: learning the structure, the parameters given a structure, hidden variables whose values are never present in the data, and values of a variable that are sometimes missing. This review describes some current literature ....

W.L. Buntine, "Graphical models for discovering knowledge ", in Advances in Knowledge Discovery and Data Mining, U. M. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. S. Uthurasamy, Eds. MIT Press, 1995.

No context found.

Buntine, W., "Graphical Models for Discovering Knowledge", in U. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthurusamy, editors, Advances in Knowledge Discovery and Data Mining, pp 59-82. AAAI/MIT Press, 1996.

No context found.

Buntine, W.: Graphical Models for Discovering Knowledge. In: Fayyad, U.M., PiatetskyShapiro, G., Smyth, P., Uthurusay, R. (eds.): Advances in Knowledge Discovery, AAAI Press / The MITPress (1996) 59-82

No context found.

W. Buntine, Graphical models for discovering knowledge, in: Advances in Knowledge Discovery and Data Mining, U.M. Fayyad, G. Piatetsky-Shapiro, P. Smyth and R. Uthurusamy, eds, 1995, pp. 59--82.

No context found.

Forthcoming. Buntine, W. 1996. Graphical Models for Discovering Knowledge. In Advances in Knowledge Discovery and Data Mining, eds. U. Fayyad, G. PiatetskyShapiro, P. Smyth, and R. Uthurusamy, 59--82.

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