K. Ali, and M. Pazzani, "Classification using Bayes Averaging of Multiple, Relational Rulebased Models", 5 th International Workshop on AI & Statistics, Ft. Lauderdale, FL, 1995. Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
Integrating Multiple Classifiers In Visual Object Detectors.. - Jaimes, Chang (2000) (1 citation) (Correct)
....strategy. Some of the most common classifier combination strategies for classifier fusion are based on rules (for details see [9] Maximum rule . Minimum rule . Median rule . Majority vote rule Other strategies include the use of decision theoretic approaches (i.e. using Bayes rule [2]) among others. In the context of our framework, classifier fusion occurs at each node of the object definition hierarchy. For each class (e.g. sand region in figure 4) several classifiers are combined during classification. For example, the decision on whether a particular region is a sand ....
K. Ali, and M. Pazzani, "Classification using Bayes Averaging of Multiple, Relational Rulebased Models", 5 th International Workshop on AI & Statistics, Ft. Lauderdale, FL, 1995.
Bayesian Integration of Rule Models - Domingos (Correct)
....form for the relevant sums (or integrals) is known. A plausible approximation is then to only attempt to find several of the most probable models, and average over these. Buntine (Buntine, 1990) has successfully applied this approach to decision tree induction (see also (Oliver Hand, 1995) Ali and Pazzani (1996) applied it to rule induction, and found that it often reduced error relative to the single best rule set. This article investigates the relationship between BMA and other multiple model methods in the context of rule induction. Specifically, it starts from the following hypothesis: 1. Ad hoc ....
....compared with BMA on eight of the larger databases in the UCI repository. BMA was applied using Equations 3 and 4 above, with each region being the region of instance space won by each rule in the model, and a uniform prior on rule models. This approach is similar to that of (Buntine, 1990) and (Ali Pazzani, 1996) . 2 An uninformed prior like the uniform is appropriate here, since all rule sets are induced in a similar way from randomly selected examples, and there is thus no a priori reason to suppose one will be more accurate than another. In the shuttle domain, the pre defined training and test sets ....
[Article contains additional citation context not shown here]
Ali, K., & Pazzani, M. (1996). Classification using Bayes averaging of multiple, relational rule-based models. In D. Fisher & H.-J. Lenz (Eds.), Learning from Data: Artificial Intelligence and Statistics V (pp. 207--217). New York, NY: Springer-Verlag.
Bayesian Model Averaging in Rule Induction - Domingos (1997) (1 citation) (Correct)
....is then to only attempt to find several of the most probable models, and average over these. Buntine [4] has successfully applied this approach to decision tree induction (see also [16] Applications of BMA to rule induction have been carried out by Kononenko [10] and by Ali and Pazzani [1]. The latter found that it often improved accuracy relative to using the single best rule set. In this paper we compare BMA with ad hoc methods for combining multiple rule sets. 2 Bayesian Averaging of Rule Sets RISE [7] is a rule induction system that assigns each test example to the class of ....
....predicts, with a weight equal to the training set accuracy 1 of the rule that won the example in that rule set [8] This ad hoc approach was compared with BMA on eight of the larger databases in the UCI repository[14] 2 . BMA was applied in the following form, very similar to that of [4] and [1]. Let n be the sample size, x the training examples, c the corresponding class labels, and H the set of models induced (i.e. each element h of H is a rule set) Then, by Bayes s Theorem, and assuming the examples are drawn independently: 1 With the Laplace correction [15] 2 Credit, ....
K. Ali and M. Pazzani. Classification using Bayes averaging of multiple, relational rule-based models. In D. Fisher and H.-J. Lenz, editors, Learning from Data: Artificial Intelligence and Statistics V, pages 207--217. Springer-Verlag, New York, NY, 1996.
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