"... In PAGE 12: ... In the appendix, we present some additional experimental results. Table4 summarizes the sensitivity of the running time of SEuS to the support threshold parameter. (Note that in worst case, the size of the output grows to a size exponential in the size of the input database as threshold is lowered; thus it is unavoidable that all methods that produce a complete output, such as SEuS, will experience a rapid rise in running time with falling thresholds.... ..."

"... In PAGE 12: ... In the appendix, we present some additional experimental results. Table4 summarizes the sensitivity of the running time of SEuS to the support threshold parameter. (Note that in worst case, the size of the output grows to a size exponential in the size of the input database as threshold is lowered; thus it is unavoidable that all methods that produce a complete output, such as SEuS, will experience a rapid rise in running time with falling thresholds.... ..."

"... In PAGE 8: ... Performance comparison and feature discovery Our miner (eFP) extends the FPgrowth* method [19] to identify the potential correlation from AMPK regulation. Table7 presents a performance comparison between our miner and algorithms FPgrowth [13], dEclat [20] and MAFIA [21]. In the comparison, we use a synthetic sparse dataset T40I10D100K http://www.... ..."

"... In PAGE 8: ...2.2 Time-sensitive effectiveness Table4 shows the average precision values at different recall levels for the three search systems at cutoff 10 when p = 0.5 using the scheme described in Section 4.... In PAGE 8: ...2. Figure 1 shows the graph representation of Table4 . We can see that AllInOneNews has an overall much better performance than Mamma News and Google News for this measure.... In PAGE 10: ... 7. The experimental results reported in Table4 suggest that AllInOneNews is more capable of getting fresher news than both Google News and Mamma News. This is consistent with our experience unrelated to this particular comparative study.... ..."

"... In PAGE 7: ...1 Preprocessing The preprocesing data mining techniques to get gene sub- sets we applied in this experiment include frequent item set and maximal frequent item set mining method with di er- ent support, K-mean, SOM, PCA, and Hierarchical cluster- ing methods. Table2 shows the size of gene sets obtained using frequent itemset and maximal frequent itemset min- Table 2: Size of gene sets obtained using frequent itemset and maximal frequent itemset with di erent support and execution time (in seconds) of frequent itemset mining(Tf), GGMs (Tg) and loglinear modeling (Tl)... In PAGE 7: ... We can see the size of frequent item set and maximal frequent item set under low support values is large. Table2 also shows the execution time of preprocessing (fre- quent itemset mining by Apriori) and that of GGMs and loglinear modeling over all subsets. We can see the exe- cution time of GGMs is trivial as it is even less than that of frequent itemset mining.... ..."

"... In PAGE 6: ...1.2 and 1.2 as being neither over-expressed nor under- expressed. Table1 shows the size of gene sets obtained using frequent itemset mining method on both Yeast and E. Coli data sets with different support values.... In PAGE 7: ...n Refs. [33,34]. For example, there are only five out of 182 E. Coli frequent itemsets (as shown in Table1 using support value 18%), which match some known complex. The reason has twofold.... ..."

"... In PAGE 3: ...11 demonstrate up to a 1 000 000-fold speedup for a taxonomy of depth seven. 2 Related Work Table1 shows a structured view of selected work on mining frequent itemsets. The table entry (frequent itemsets, ordinary) shows earlier work on mining fre- quent itemsets (Subsection 2.... In PAGE 3: ...3. Table1 high- lights that there is a blank entry with no existing work: to mine max frequent g-itemsets. This paper fllls that blank.... ..."