David D. Lewis. Naive Bayes at forty: The independence assumption in information retrieval. In ECML-98: Proc. Tenth European Conf. Machine Learning, pages 4--15, Springer, Berlin, April 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....classification accuracy. It is particularly attractive for interactive applications due to the speed with which it can be applied. This speed is derived from its use of a simplifying attribute independence assumption. Naive Bayes has a long history of application in information retrieval [12] and has gained some popularity in the machine learning community. It has numerous desirable features. It is extremely e#cient. It is provably optimal bar only for two explicit assumptions, the attribute independence assumption and the assumption that the estimates based on frequency are accurate. ....

D. D. Lewis. Naive Bayes at forty: The independence assumption in information retrieval. In ECML-98: Proceedings of the Tenth European Conference on Machine Learning, pages 4--15, Chemnitz, Germany, April 1998. Springer.

....belonging to 10 Yahoo categories. All the categories considered are subcategories of the category Science. We performed our experiments using two different classifiers to verify the robustness of the techniques aside from the algorithm used. We used a probabilistic classifier called Naive Bayes [1] and a perceptron based classifier using kernel functions called Kernel Perceptron [2] or, simply, Perceptron) We used a linear kernel in all the experiments except the one of Section 7. The results of the experiments are averages of 6 random test training splits of the dataset. The evaluation ....

D. D. Lewis, "Naive (Bayes) at forty: The independence assumption in information retrieval", Proceeding of ECML-98, 10th European Conference on Machine Learning, Springer Verlag, Heidelberg, DE, 1998, pp. 4-15.

....to appear in classification benchmarks outperformed by other, more recent, methods. Despite this fate, in the past few years this simple technique has emerged once again, basically due to its results both in performance and speed in the area of information retrieval and document cate gorization [1, 2]. Recent experiments on benchmark databases have also shown that Naive Bayes outperforms several standard classifiers even when the independence assumption is not met [3] Additionally, the statistical nature of Naive Bayes implies interesting theoretic and predictive properties and, if the ....

Lewis, D.: Naive bayes at forty: The independence assumption in information retrieval. In N'edellec, C., Rouveirol, C., eds.: Proceedings of ECML-98, 10th European Conference on Machine Learning. Volume 1398.25., Springer Verlag, Heidelberg, DE (1998) 4-15

.... words: Log linear modelling, Text Classification, Naive Bayes, Multinomial, Smoothing, Length Normalization I Introduction The naive Bayes text classifier has long been a core technique in information retrieval and, more recently, it has emerged as a focus of research itself in machine learning [1]. Recent work in this area has focused on its probabilistic analysis and, in accordance with [2] we can identify two basic instantiations of this classifier: the Bernoulli model and the multinomial model. The Bernoulli model works with documents represented as vectors of binary features ....

Lewis, D.D.: Naive Bayes at Forty: The Independence Assumption in Information Retrieval. In: Proc. of the ECML'98. (1998) 4 15

....Text classi cation is a good example. Standard text classi cation procedures developed in information retrieval are based on either binary or integer valued features. In both cases, it has been recently shown that these procedures are closely tied to more well founded statistical decision rules [6, 8]. In fact, both classical and new pattern recognition techniques are currently being tested with success on this task [6, 7, 9, 12] One of the most widely used new text classi cation techniques is the socalled naive Bayes classi er (for binary data) 6, 8, 9] It is a Bayes plug in classi ....

.... statistical decision rules [6, 8] In fact, both classical and new pattern recognition techniques are currently being tested with success on this task [6, 7, 9, 12] One of the most widely used new text classi cation techniques is the socalled naive Bayes classi er (for binary data) [6, 8, 9]. It is a Bayes plug in classi er, which assumes that the binary features are class conditional independent, and thus each pattern class can be represented as a multivariate Bernoulli distribution. While this technique is still considered to be among the most Work supported by the Spanish CICYT ....

D. D. Lewis. Naive Bayes at Forty: The Independence Assumption in Information Retrieval. In Proc. of the ECML'98, pages 4-15, 1998.

....Text classi cation is a good example. Standard text classi cation procedures developed in information retrieval are based on either binary or integer valued features. In both cases, it has been recently shown that these procedures are closely tied to more well founded statistical decision rules [6, 9]. In fact, both classical and new pattern recognition techniques are currently being tested with success on this task [6, 7, 10, 13] One of the most widely used new text classi cation techniques is the socalled naive Bayes classi er (for binary data) 6, 9, 10] It is a Bayes plug in ....

.... statistical decision rules [6, 9] In fact, both classical and new pattern recognition techniques are currently being tested with success on this task [6, 7, 10, 13] One of the most widely used new text classi cation techniques is the socalled naive Bayes classi er (for binary data) [6, 9, 10]. It is a Bayes plug in classi er, which assumes that the binary features are class conditional independent, and thus each pattern class can be represented as a multivariate Bernoulli distribution. While this technique is still considered to be among the most accurate text classi ers, it is ....

D. D. Lewis. Naive Bayes at Forty: The Independence Assumption in Information Retrieval. In Proc. of the ECML'98, pages 4-15, 1998.

....attributes of a data point are conditionally independent within each submodel, i.e. that P (x i ; x j jk) P (x i jk)P (x j jk) where x i and x j such that j 6= i are attributes of x. While typically not true, this assumption simplifies probability estimation and often works well in practice (Lewis, 1998). For a d dimensional data point and a given class k we have P (xjk) d Y i=1 P (x i jk) 2) We use discrete naive Bayes submodels. The values of each attribute are placed in a small finite number of bins. A bin is an internal value used in place of the true value of an attribute. For ....

....the disk drive dataset using a second method, a naive Bayes classifier. Naive Bayes learning is a well known supervised learning method that yields a classifier distinguishing between each class of data. For further information on the standard naive Bayes classifier, the reader is referred to (Lewis, 1998). We have also applied boosting to the standard classifier to obtain an ensemble of voting classifiers (Freund, 1995; Elkan, 1997) but for this application boosting produces no appreciable improvement in results. As with naive Bayes EM, with the naive Bayes classifier we place continuous ....

Lewis, D. (1998). Naive Bayes at forty: The independence assumption in information retrieval. Conference proceedings of European Conference on Machine Learning (pp. 4--15).

No context found.

David D. Lewis. Naive Bayes at forty: The independence assumption in information retrieval. In ECML-98: Proc. Tenth European Conf. Machine Learning, pages 4--15, Springer, Berlin, April 1998.

No context found.

David D. Lewis. Naive Bayes at forty: The independence assumption in information retrieval. In Proceedings of ECML-98, The Tenth European Conference on Machine Learning, pages 4--15, 1998.

No context found.

D. Lewis. Naive bayes at forty: The independence assumption in information retrieval. In Proc. 10th European Conference on Machine Learning ECML-98, pages 4--15, 1998.

No context found.

D. Lewis. Naive bayes at forty: The independence assumption in information retrieval. In Proc. 10th European Conference on Machine Learning ECML-98, pages 4--15, 1998.

No context found.

Lewis, D., 1998. Naive Bayes at forty: The independence assumption in information retrieval. Conference proceedings of European Conference on Machine Learning (pp. 4--- 15).

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