### Table 1. Synthetic Microarray Datasets

"... In PAGE 3: ... The synthetic data used is microarray- like data with explicitly modeled overlapping clusters generated as suggested by Banerjee et al. in [1] with a clear ground truth, described in Table1 . The real dataset used is the Gasch dataset [6] consisting of the 995 most actively expressed genes of yeast under 173 environmental stress conditions.... ..."

### Tables Table 1. Main results of geometry optimisations. zeolite cluster zeolite cluster change without acetonitrile with acetonitrile after absorption

### Table 1. Comparison of the imputation models in the LDPE plant for two simulated failures. Absolute mean errors and statistical properties of the imputed dataset (mean, median and variance) for three different levels of missingness. Imputation

"... In PAGE 4: ... The data used to train the system included all the 148 variables, with both complete and incomplete patterns. Table1 summarizes the results obtained using the ... In PAGE 6: ... Also, in each case the effect of the amount of missing data has been studied, varying the missing data ratio from 10% to 70% of the available patterns. Inspection of Table1 leads to several observations: (i) Regardless of the methods used to construct the individual single imputation systems the absolute mean error of the aggregated response of an ensemble of maps is always lower than that of each individual system; (ii) all imputation systems maintain an stable behavior with the increase of missing data, independently of the imputation technique used. The imputation systems based on the SOM are very robust with respect to the amount of missing data.... ..."

### Table 1. Microarray datasets

2005

"... In PAGE 4: ...l., 2005; Van Driessche et al., 2002) were weighted based on the number of replications. Table1 summarizes the genotypes, phenotypes and weights. Weighted datasets were concatenated into one matrix of 4,839 non-redundant genes and 186 instances.... In PAGE 7: ... The trajectories of many of these genes have been confirmed by Northern blots or by comparison to published data (Supplement Table 3). To refine the clusters, we considered data from 13 mutant strains ( Table1 ). We concatenated the mutant datasets, performed cluster analysis in each of the 17 modes and found that 14 modes were divided into significant subgroups, resulting in 57 sub-modes ... ..."

Cited by 1

### Table 12 Related woks on DNA microarray data

2005

"... In PAGE 7: ...3. Related works Studies about cluster validation of the DNA microarray data are shown in Table12... ..."

### Table 1: Summary of the Myrtveit et al. Comparison of Imputation Techniques [1]

"... In PAGE 3: ...Table 1: Summary of the Myrtveit et al. Comparison of Imputation Techniques [1] Table1 summarises the results Myrtveit et al. obtained when trying to assess effort prediction models derived by linear regression.... In PAGE 9: ...2943553378.5% Table1 0: Comparison of Regression Model Accuracy for Multi-National Data Set Table 10 shows the results of using a stepwise regression procedure to build prediction models for EFFORT using the raw data without imputation, with imputed values using SMI and kNN imputation. No model could be built for the NI data, however, when using imputation methods the picture changes somewhat.... ..."

### TABLE 1. SELECTED GENES FROM CLUSTERING ANALYSIS OF OLIGONUCLEOTIDE MICROARRAY DATA

### Table 2: Software tools for clustering-based analysis of microarray data

### Table 1: Statistics that can be used to evaluate an incomplete factorization.

1997

"... In PAGE 3: ... To try to understand what can happen in an incomplete factorization, a number of statistics can be monitored. These statistics, shown in Table1 , can be monitored during the course of a factorization, or after the factorization has been computed.... In PAGE 15: ...15 We begin by showing how the statistics presented in Table1 can be used to determine what di culties are arising when an incomplete factorization fails. Table 3 lists problems that could not be solved using ILU(0) as a preconditioner, along with their values of the statistics, and the causes of failure as classi ed by Figure 1 and the comments in Section 1.... In PAGE 16: ... Without the shift, we could only solve problems FIDAP006, FIDAPM02, and FIDAPM08. In Table1 0(a) we perform a parameter study of the e ect of changing thresh for the... In PAGE 17: ... Thus we can use block ILU with a block size of 4, and illustrate the use of a block shift (9). Table1 0(b) shows the results. Here, thresh is the ratio of the largest singular value to the smallest in (9).... In PAGE 19: ... His conclusion also was that increasing the accuracy does not seem to help, unless of course, the accuracy approaches that of a direct solve. Table1 2(a) shows `condest apos; for BILUK applied to the PULLIAM1 problem. The values are much smaller.... In PAGE 19: ... An alternative when an threshold-based ILU is unstable is to use a level-based factorization. Table1 2(b) shows the number of GMRES steps required to solve the PULLIAM1 problem. Table 13(a) shows the number of GMRES steps required to solve the BBMAT problem, along with timings on a Cray-C90 computer.... In PAGE 19: ... Table 12(b) shows the number of GMRES steps required to solve the PULLIAM1 problem. Table1 3(a) shows the number of GMRES steps required to solve the BBMAT problem, along with timings on a Cray-C90 computer. Note that block size 16 is fastest, even though the factorization requires 50 percent more storage than block size 8 (due to some explicit storage of zeros; see Table 13(b)), partially due to better vectorization.... In PAGE 19: ... Table 13(a) shows the number of GMRES steps required to solve the BBMAT problem, along with timings on a Cray-C90 computer. Note that block size 16 is fastest, even though the factorization requires 50 percent more storage than block size 8 (due to some explicit storage of zeros; see Table1 3(b)), partially due to better vectorization.... In PAGE 28: ...0 4.25e+02 y (b) Block ILU(0) Table1 0: Stablized ILU(0), (a) pointwise, and (b) block versions, for the WIGTO966 problem. matrix method max(L+U) 1/pivot condest steps lhr01(t) ILUTP(30) 4.... In PAGE 28: ...06e+02 5.56e+06 37 Table1 1: Solution of some harder problems. Notes: (t) transposed data structure and ILUTP... In PAGE 29: ...2 21.9 (b) Number of scalar nonzeros for BILUK, in millions Table1... ..."

Cited by 55

### Table 1. Microarray Data Storage and Exchange Systems

"... In PAGE 24: ... The development of new technologies to store digital information are transforming the life sciences and enabling scientists to record vast quantities of data. These advances and the improvement in the sensitivity of microarray technology have motivated the development of a considerable number of specialized databases ( Table1 ). As the relevance of microarray experiments increases, the use of this technology for diagnostics and clinical research present a new paradigm in the storage of this information.... ..."