### Table 5: Max Average F-Score: Multinomial and Complement Naive Bayes Smoothing Parameter Multinomial Naive Bayes Complement Naive Bayes

### Table 2. F1 measures for all models (Naive Bayes, EM, and Hierarchical Dirichlet) on the eight benchmark tax- onomies. The Hierarchical Dirichlet results are presented both for the best smoothing parameter , and for a xed for all datasets.

2005

"... In PAGE 7: ... tors were also used to initialize both the standard EM and our regularized EM algorithms. In Table2 we present the classi cation results of the di erent approaches on the eight datasets. We observe that the standard EM algorithm performs poorly on several datasets.... In PAGE 7: ... Our smoothed variant of the EM algorithm per- forms signi cantly better than both the standard EM and the Naive Bayes classi er on almost all datasets. In Table2 we present the classi cation results for the best value of on each dataset as well as the results with a xed = 2. We observe that even when the chosen is not the best for the dataset, smoothed es- timates of the classi cation parameters result in con- siderably better accuracy than without smoothing.... ..."

Cited by 7

### Table 1. Accuracy for each method; NB corresponds to the naive Bayes classi- fier, NB+EM corresponds to the naive Bayes classifier enhanced with EM, and SVM+NB+EM corresponds to the SVM that uses the Fisher kernel extracted from NB+EM model.

2006

"... In PAGE 9: ... 5.1 Results Comparison of methods Table1 shows the accuracy values for the various methods. Incorporation of unlabeled data improves classification accuracy of the naive Bayes classifiers for this task.... ..."

Cited by 2

### Table 13. Properly classifled examples: C4.5 (left) Naive Bayes (right) DB N

### Table 3 Performance of the naive Bayes classifier, summed over all 10 test runs.

2003

"... In PAGE 11: ... This interpolated probability model was used in order to smooth the probability distribution, avoiding the problems that can arise if certain feature-value pairs are assigned a probability of zero. The performance of the naive Bayes classifier is summarized in Table3 . For each of the 10 test sets of 89 items taken from the corpus, the remaining 804 of the total 893 sentences were used to train the model.... ..."

Cited by 4

### Table 2 summarizes the overall results of information extraction for proper nouns.

"... In PAGE 19: ... Table2 . The Overall Results of Information Extraction for People Finding Information Extraction # (no.... ..."

### Table 1: Model information and timings, in seconds on 195Mhz R10k for naive and hierarchical silhouette extraction methods un- der an orthographic view.

1999

"... In PAGE 5: ... However the software methods provide more flexibility and, potentially, better performance. Table1 presents the information of two extreme cases. These cases are based on orthographic views.... ..."

Cited by 70

### Table 1: Predictive accuracy for the learned models alone (naive Bayes), the operon map made from the predictions of the learned models, and for randomly chosen operon maps.

2000

"... In PAGE 8: ...iven test case (i.e., a known promoter or terminator), that we would not have in the case of a currently undis- covered operon. Table1 shows the overall accuracy rates for this ex- periment, as well as the false positive and true positive rates. The false positive (FP) rate is de ned as FP FP+TN , and the true positive (TP) rate is de ned as TP TP+FN .... In PAGE 9: ... Table 2 shows the accuracy of operon maps made using models that consist of sin- gle feature groups, and Table 3 shows the accuracy of operon maps made using models that leave one feature group out. For reference, both tables also include the result from Table1 for the naive Bayes models that use all features. Tables 2 and 3 illustrate several interesting points.... In PAGE 9: ... An ROC curve shows the re- lationship between the true positive and false positive rates as we vary a threshold on the con dence of our predictions. For example, the results in the rst row in Table1 were determined by treating our naive Bayes model as a classi er; when the posterior probability of a candidate operon was greater than 0.5, we classi ed it as an operon.... In PAGE 9: ... We get a single ROC curve from our 10 learned models by pooling the test-set predictions of these models. This curve is informative since, unlike the overall accuracy numbers in Table1 , it does not depend on prior probabilities of the two classes or any particular misclassi cation costs (this is a prop- erty of ROC curves). Figure 4 also illustrates that our learned models have considerable predictive value.... ..."

Cited by 24

### Table 1: Predictive accuracy for the learned models alone (naive Bayes), the operon map made from the predictions of the learned models, and for randomly chosen operon maps.

2000

"... In PAGE 8: ...iven test case (i.e., a known promoter or terminator), that we would not have in the case of a currently undis- covered operon. Table1 shows the overall accuracy rates for this ex- periment, as well as the false positive and true positive rates. The false positive (FP) rate is de ned as FP FP+TN , and the true positive (TP) rate is de ned as TP TP+FN .... In PAGE 9: ... Table 2 shows the accuracy of operon maps made using models that consist of sin- gle feature groups, and Table 3 shows the accuracy of operon maps made using models that leave one feature group out. For reference, both tables also include the result from Table1 for the naive Bayes models that use all features. Tables 2 and 3 illustrate several interesting points.... In PAGE 9: ... An ROC curve shows the re- lationship between the true positive and false positive rates as we vary a threshold on the con dence of our predictions. For example, the results in the rst row in Table1 were determined by treating our naive Bayes model as a classi er; when the posterior probability of a candidate operon was greater than 0.5, we classi ed it as an operon.... In PAGE 9: ... We get a single ROC curve from our 10 learned models by pooling the test-set predictions of these models. This curve is informative since, unlike the overall accuracy numbers in Table1 , it does not depend on prior probabilities of the two classes or any particular misclassi cation costs (this is a prop- erty of ROC curves). Figure 4 also illustrates that our learned models have considerable predictive value.... ..."

Cited by 24

### Table 1: Predictive accuracy for the learned models alone (naive Bayes), the operon map made from the predictions of the learned models, and for randomly chosen operon maps.

2000

"... In PAGE 8: ...iven test case (i.e., a known promoter or terminator), that we would not have in the case of a currently undis- covered operon. Table1 shows the overall accuracy rates for this ex- periment, as well as the false positive and true positive rates. The false positive (FP) rate is defined as FP FP+TN , and the true positive (TP) rate is defined as TP TP+FN .... In PAGE 9: ... Table 2 shows the accuracy of operon maps made using models that consist of sin- gle feature groups, and Table 3 shows the accuracy of operon maps made using models that leave one feature group out. For reference, both tables also include the result from Table1 for the naive Bayes models that use all features. Tables 2 and 3 illustrate several interesting points.... In PAGE 9: ... An ROC curve shows the re- lationship between the true positive and false positive rates as we vary a threshold on the confidence of our predictions. For example, the results in the first row in Table1 were determined by treating our naive Bayes model as a classifier; when the posterior probability of a candidate operon was greater than 0.5, we classified it as an operon.... In PAGE 9: ... We get a single ROC curve from our 10 learned models by pooling the test-set predictions of these models. This curve is informative since, unlike the overall accuracy numbers in Table1 , it does not depend on prior probabilities of the two classes or any particular misclassification costs (this is a prop- erty of ROC curves). Figure 4 also illustrates that our learned models have considerable predictive value.... ..."

Cited by 24