### Table 4: Comparison with the model of Cox and Chu (ppb)

"... In PAGE 24: ... In our fitting of their model for the data of Research Triangle area for 1993- 1997, all the meteorological factors they used were included except mixing height and opaque cloud cover, which are not available to us. The fitting results of their model are distinctly worse than the presented model, as shown in Table4 . On the basis of the simplest measure, mean squared error of prediction, the Cox and Chu [1993] model (13.... ..."

### Table 1: The summarized characteristics of UC and CHU traces.

"... In PAGE 4: ... Take the UC trace as an example again. As shown in Table1 , we will at best have 45% of the hit ratio with a cache of size 7.2 GB in the extreme case where the requested objects are not stored in disk cache in the first time.... In PAGE 8: ... For HR and BHR, we plot the relative ratios to that in a cache of infinite size. As shown in Table1 , the HR and BHR are 0.... ..."

### Table 1. Results for the Abalone dataset. We report also the unbiased standard deviation measured on the 10 different shuffles of the dataset. SVR Chu refers to [Chu et al., 2004].

"... In PAGE 7: ... Hyperparameters selection and final evaluation were performed using the same procedure adopted for SVR. The results for the Abalone dataset, both for the training set (tr) and the test set (ts), are shown in Table1 where we report also the results obtained for SVR in [Chu et al., 2004].... ..."

### Table 2: Number of genes matching to each pro le for (a) Chu et al. clustering method,

in Bootstrapping Cluster Analysis: Assessing the Reliability of Conclusions from Microarray Experiments

2001

"... In PAGE 6: ... Gene g is assigned to pro le p if rgp gt; 0:9 and rgp is larger than rgq for the remaining 6 pro les q. From columns (a) and (b) in Table2 , we see that the number of genes clustering to each pro le is somewhat larger here than for Chu et al., except for pro le 2 (Early I).... In PAGE 8: ...Table2 shows the much smaller numbers of genes that prove to be reliable matches to the clusters at the 95% level. We refer to these as stable genes.... ..."

Cited by 53

### Table 2. Advantages and disadvantages of various modes of communication - adapted from Chu and Chan (1998).

"... In PAGE 13: ... The options of print, voice/video (classroom), CD-ROM, and Internet exist. All have their own advantages and disadvantages ( Table2 ). Future courses are not likely to be delivered solely in one medium.... ..."

### Table 2: Average percentage gaps and standard deviations of the di erent approaches, using the whole Chu and Beasley 500-variables instance set.

2005

"... In PAGE 9: ... The neighborhoods are ordered as follows: N1:=SWP(xf), N2:=RGF (xf; 1), N3:=AGR(xf; 1), N4:=RGF (xf; 2), N5:=AGR(xf; 2). In Table2 we show results of the following algorithm variants: VNS with neighborhoods N1 to N3 only (VNS{N1 3), VNS with neighborhoods N1 to N5 only (VNS{N1 5), VNS with the ILP-based neighborhoods (VNS), RGVNS with the ILP-based neighborhoods (RGVNS), RGVNS with additionally N1 to N3, and RGVNS with additionally N1 to N5. Each algorithm was given a run-time of 500 seconds.... ..."

Cited by 1

### Table 1. Summary of the experimental data collected by Chu and his colleagues (1998). The table contains average relative expression ratios after application of a natural log transform. Time

2000

"... In PAGE 4: ...oints (0hrs, 0.5hr, 2hrs, 5hrs, 7hrs, 9hrs, 11.5hrs) during sporulation. Thus, the matrix to be analyzed has 6118 rows of genes and 7 columns of conditions corresponding to each of the measured time points. Table1 reports the mean, median, and variance of each time point from the sporulation data over all genes. The means and medians are slightly negative but quite close to zero.... ..."

Cited by 70

### Table 1. Summary of the experimental data collected by Chu and his colleagues (1998). The table contains average relative expression ratios after application of a natural log transform. Time

2000

"... In PAGE 4: ...0hrs, 0.5hr, 2hrs, 5hrs, 7hrs, 9hrs, 11.5hrs) during sporulation. Thus, the matrix to be analyzedhas 6118 rows of genes and 7 columns of condtions corresponding to each of the measured time points. Table1 reports the mean, median, and variance of each time point from the sporulation data. The means and medians are slightly negative but quite close to zero.... ..."

Cited by 70

### Table 1. Summary of the experimental data collected by Chu and his colleagues (1998). The table contains average relative expression ratios after application of a natural log transform. Time

2000

"... In PAGE 4: ...oints (0hrs, 0.5hr, 2hrs, 5hrs, 7hrs, 9hrs, 11.5hrs) during sporulation. Thus, the matrix to be analyzed has 6118 rows of genes and 7 columns of conditions corresponding to each of the measured time points. Table1 reports the mean, median, and variance of each time point from the sporulation data over all genes. The means and medians are slightly negative but quite close to zero.... ..."

Cited by 70

### Table 2. Zero-one error of NNRank, SVM, GP-MAP, and GP-EP on the eight datasets. SVM denotes the support vector machine method (Shashua amp; Levin, 2003; Chu amp; Ghahramani, 2005a). GP-MAP and GP-EP are two Gaussian process methods using Laplace approximation (MacKay, 1992) and expectation propagation (Minka, 2001) respectively (Chu amp; Ghahramani, 2005a). The results are the average error over 20 trials along with the standard deviation. We use boldface to denote the best results.

704