### Table 2: Related Work organized along three dimensions. Note that SWAP results subsume INT results, which in turn subsume EXT results. However, many of these results are more general than their entry in this table suggests. For example, the framework of Herbster and Warmuth (1998) deals with time-varying experts as well as external regret. Also, an appropriate bound on distribution-regret can imply no-regret: i.e., convergence to zero of action regrets (see the discussion of the Hoeffding-Azuma lemma in Section 5.2).

### Table 3. Hyperparameters for estimating the parameters of mixture of normal distributions. Gibbs Sampling Quasi-Bayesian MLE

"... In PAGE 38: ... 33 The hyperparameter settings used in estimation procedures are given in Table3 . The hyperparameters are chosen so that the priors are roughly of the same magnitude with theoretical parameters; however, only a small weight is assigned to the prior information to reflect vague prior knowledge.... ..."

### Table 1: Average number of genes out of the 120 with the smallest p-values identified in common based on analyzing sub- samples of the IHF data set.

"... In PAGE 12: ... With the exception of the two statistics based on ratios of means (standard deviation for the ratio of raw means for 2 by 2, ratio of raw means for 3 by 3, ratio of mean of logs for 2 by 2 and ratio of mean of logs for 3 by 3 were 25, 32, 26, and 13 respectively), standard deviations over comparisons were generally small and between 7 and 13. The different combinations of window size and hyperparameter seemed to have little effect on the consistency of the Bayesian approach so only outcomes corresponding to the best and worst parameter combinations are presented in Table1 . The comparisons among statistical approaches presented in Table 2 were generated in a similar manner to those described above.... In PAGE 18: ... Similarly, in the case of the comparisons of size three, 36 measures of consistency were made from comparisons having one replicate in common. Table1 summarizes the consistency of the different statistical approaches. As might be expected additional replications of an experiment result in greater consistency at identifying the same set of genes as being up- or down-regulated.... In PAGE 20: ...The Bayesian approach allows the identification of more true positives with fewer replicates The data in Table1 show that additional experimental replications result in the identification of a more consistent set of up- or down- regulated genes, and that the Bayesian statistical approach identifies a more consistent set than a simple t-test. The natural question that arises is whether these genes are true positives.... ..."

### Table 1: Carbon Emissions in Post- Planned Economies, 1996 ( million tons of carbon)

1997

"... In PAGE 3: ... Prospects for A Russian No-Regrets Strategy No aspect of global economic development will have more impact on greenhouse gas emissions than the manner in which transition economy energy use evolves over the next two decades. Russia represents half the emissions of these nations (see Table1 ), but has at its disposal a number of tools to reduce emissions growth. Available mitigation options in Russia include: C Energy efficiency C Switching to natural gas C Nuclear power C Renewable energy... ..."

### Table 1: Average misclassification rate R and the standard deviation of the mean at optimal hyper-parameters C, S and .

### Table 3. Bayesian combination methods

"... In PAGE 4: ...Thereject results of a classifier were used in finding the optimal product set. The five classifiers, shown in Table 2, were evaluated by the Bayesian combination methods abbreviated as in Table3 and the BKS method. From the Figure 2, the second-order dependency provides higher performance than the first-order dependency, however, the third-order depen- dency does not provide higher performance than the second- order dependency in all groups.... ..."

### Table 1 shows different optimal actions when different probabilities of Inventory satisfaction at state S = (OR=Yes, OV=Yes, IA=Unknown, PSA=Unknown, SMA=Unknown, GR=Unknown, SS=Unknown, GS=Unknown) are used in the SMDP model.

"... In PAGE 7: ...50 Check Preferred Supplier 0.42 Check Preferred Supplier Table1 : Optimal policy changes when the probability of inventory satisfaction changes Figure 3: Optimal policy responses to a Web Service failure (a) and an unexpected Web Service invocation result (b). Figure 3 is realized in the event of a Web service failure and an unexpected Web service response is obtained.... ..."

### Table 2. Hyperparameter Specifications

in runoff

"... In PAGE 5: ... The fixed constants tij are listed in Table 2. We assume that the si 2 are independent, and have inverse gamma distributions, with parameters ai and bi, also given in Table2 . (The quantities tij, ai, and bi are often called hyper- parameters.... In PAGE 11: ... Also, for i 5 0 and 2, define Yi 5 diag[ti1, ti2], where diag [ ] represents a diagonal matrix whose elements are given by the indicated vector. (Values of tij are specified in Table2 .) Similarly, let Y1 5 diag [t11, t12, t13].... ..."

### Table 3: Networks In Bayesian Network Repository

"... In PAGE 7: ... We were able to compute the treewidth of some graphs whose optimal treewidth was not yet known. The results are shown in Table3 . The time-bound used was 1 hour.... ..."