### Table 4. Comparison between SVM, JTSVM, rTSVM and DA (all with quadratic hinge loss (l2)). For each method, the top row shows mean error rates with model selection; the bottom row shows best mean error rates. u/t denotes error rates on unlabeled and test sets. Also recorded in performance of DA with squared loss (sqr). usps2 coil6 pc-mac eset2

2006

"... In PAGE 7: ... This experimental setup neutralizes any undue advantage a method might receive due to di er- ent sensitivities to parameters, class imbalance issues and shortcomings of the model selection protocol. Comparing DA, JTSVM and rTSVM Table4 presents a comparison between DA, JTSVM and rTSVM. The baseline results for SVM using only labeled examples are also provided.... In PAGE 7: ... Being a gradi- ent descent technique, rTSVM requires loss functions to be di erentiable; the implementation in (Chapelle amp; Zien, 2005) uses the l2 loss function over labeled examples and an exponential loss function over unla- beled examples. The results in Table4 for DA and JTSVM were also obtained using the l2 loss. Thus, these methods attempt to minimize very similar ob- jective functions over the same range of parameters.... In PAGE 8: ... Here, too, annealing gives signi cantly better results. Performance with Squared Loss In Table4 we see that results obtained with the squared loss are also highly competitive with other methods on real world semi-supervised tasks. This is not surprising given the success of the regularized least squares algorithm for classi cation problems.... ..."

Cited by 8

### Table 2: Results on Small-Scale Datasets. We report the best test error over the hyperparameters of the algorithms, as in the methodology of Chapelle and Zien (2005). SVMLight-TSVM is the implemen- tation in SVMLight, rTSVM is the primal gradient descent method of Chapelle and Zien (2005) and CCCP-TSVM is our method. CCCP-TSVM with s = 0 is our method using the Symmetric Hinge Loss (Figure 1, left). The latter does not perform as well as the same loss function with a plateau, the Symmetric Ramp Loss (Figure 1, right), termed CCCP-TSVM.

2006

Cited by 13

### Table 1. Semi-supervised Learning with Deterministic An- nealing.

2006

"... In PAGE 5: ... The parameter T is decreased in an outer loop until the total entropy falls below a threshold. Table1 outlines the steps for the algorithm with default parameters. In the rest of this paper, we will abbreviate our method as DA (loss) where loss is l1 for hinge loss, l2 for quadratic hinge loss and sqr for squared loss.... ..."

Cited by 8

### Table 1. Semi-supervised Learning with Deterministic An- nealing.

2006

"... In PAGE 5: ... The parameter T is decreased in an outer loop until the total entropy falls below a threshold. Table1 outlines the steps for the algorithm with default parameters. In the rest of this paper, we will abbreviate our method as DA (loss) where loss is l1 for hinge loss, l2 for quadratic hinge loss and sqr for squared loss.... ..."

Cited by 8

### Table 3: Mean (standard deviation) of weighted ( = 0:01) error rate values on the IJCNN dataset. C+ = C, C = C C+ = C, C = C C+, C tuned Full Weighted Hinge

in Abstract

"... In PAGE 8: ... We take = 0:01. The top half of Table3 reports weighted error rates associated with validation and test. The weighted hinge loss model performs best.... In PAGE 8: ...The presence of the threshold parameter 0 is important for the rst three methods. The bottom half of Table3 gives the performance statistics of the methods when threshold is not tuned. Interestingly, for the weighted hinge loss method, tuning of threshold has little effect.... ..."

### Table 1. Comparison of the average generalization error and standard deviation , computed over the 10 splits, for ve different algorithms on ve UCI binary datasets. The results quoted for SVMt are taken from ([7]).

"... In PAGE 3: ...ttp://ida.first.gmd.de/ raetsch. Each dataset comprises the rst 10 splits of the 100 split available, where each split has a training and a test set. In Table1 we report a comparison of SVM with Hinge loss and RBF kernel ( Ki;j = e Kjj xi xjjj2 ), QLYSVM with a RBF kernel used both for K and S (with hyperparameter S), and QLYSVM (referred as QLYSVMe) where S is the exponential matrix eK, being K the previously de ned RBF kernel matrix. The last two algorithms were evaluated also employing a normalized S matrix (QLYSVMN and QLYSVMeN).... ..."

### Table 1. Analyses of body mass index, gender, and age in controls from Genair/EPIC and in the EPIC cohort EPIC*, c (b) Controls from

"... In PAGE 7: ... It is possible that cross- sectional results in controls will produce a quite biased representation of associations in the overall cohort at baseline. The issues of bias and precision are illustrated in Table1 . The association between body mass index and gender and age was analyzed in the controls from Genair/EPIC and the corresponding portion of the EPIC cohort in which Genair/ EPIC was nested.... ..."

### Table 2 Regression coefficients describing the logistic regression model for Tanga Region, Tanzania*. LST land surface tem- perature; NDVI normalized difference vegetation index

"... In PAGE 4: ... Results A number of logistic regression models were fitted to a 50% random sub-sample of schools in Tanga Region. Table2 presents the final model results and shows that altitude has a negative effect on the probability of a school having prevalence gt; 50%, whereas minimum LST and mean NDVI both have a positive effect. The remaining 50% of schools in Tanga Region not selected to develop the model were used to assess the accuracy of the model.... ..."

### Table 4. Stepwise Regression

2002

"... In PAGE 4: ... Tables 1, 2, 3 show the regression results for single variable models built using the the entire data set. Stepwise regressions were performed to understand the combined effects of the various variables, and are given in Table4 . Additional experiments with polynomial regression models showed poorer cor- relation than the linear models.... In PAGE 4: ... 5.2 Stepwise Regression Table4 shows the stepwise regression model. f() denotes a model of the variables.... ..."

Cited by 11