### Table 2 Real-world problems.

"... In PAGE 9: ... These problems represent (a) scheduling and bounded model checking for timed automata, (b) verification of hardware models, including a load-store unit and an out-of-order execution unit. Consider Table2 : we compare TSAT++ (full IS(2) preprocessing, prime implicants reduction and detection of the smallest reason for the post-office problem and early-pruning plus smallest reason for the others), SEP without conjunction matrix and MathSAT, on the biggest problems SEP could tackle, found in either category. In category (b) variables are restricted to take integer values, and MathSAT has been excluded from the comparison.... ..."

### Table 4. Real-world data matrix

2004

"... In PAGE 9: ...able 3. Control condition 1 data matrix.......................................................................... 43 Table4 .... In PAGE 70: ... One-tailed, paired t-tests were performed to test hypothesis 1. Accuracy, task completion, and fatigue rating of cell A1B1 in Table4 were compared with accuracy, completion rate, and fatigue of cell R1F1V1 in Table 5. Similarly, cell A2B2 in Table 4 was compared with R2F2V2 in Table 5, and A3B3 with R3F3V3 using separate t-tests at a significance of 0.... In PAGE 70: ... Accuracy, task completion, and fatigue rating of cell A1B1 in Table 4 were compared with accuracy, completion rate, and fatigue of cell R1F1V1 in Table 5. Similarly, cell A2B2 in Table4 was compared with R2F2V2 in Table 5, and A3B3 with R3F3V3 using separate t-tests at a significance of 0.05.... In PAGE 72: ....1.2 Fatigue rating One-tailed, paired t-tests revealed that context-switching had a significant effect on fatigue rating for each of the three distances. Thus, there was a significant difference between cell A1B1 in Table4 and R1F1V1 in table 5, A2B2 and R2F2V2, and A3B3 and R3F3V3 for fatigue rating. Participants found task performance in the real-world context (control condition 1) less fatiguing for all distances as compared to the augmented reality ... ..."

Cited by 2

### Table 2: Description of real-world domains

1995

"... In PAGE 4: ... All are from the UCI repository of machine learning databases [Murphy and Aha, 1994]. Table2 gives a brief summary of the domains, including the data set size, the number of binary (B), nominal (N), continuous-valued (C), total (T) attributes, and the number of classes. Phoneme and Stress are two basic subproblems of the Nettalk domain.... ..."

Cited by 14

### Table 2: Description of real-world domains

1995

"... In PAGE 4: ... All are from the UCI repository of machine learning databases [Murphy and Aha, 1994]. Table2 gives a brief summary of the domains, including the data set size, the number of binary (B), nominal (N), continuous-valued (C), total (T) attributes, and the number of classes. Phoneme and Stress are two basic subproblems of the Nettalk domain.... ..."

Cited by 14

### Table 3: Results on real-world domains. Numbers enclosed in parentheses represent stan- dard deviations.

"... In PAGE 13: ...ule hypothesis: if E then Class = c. The algorithm is detailed in Appendix B. 5.1 Experimental Methodology Table3 compares the performance of the learning algorithm described above when the search for the best expression is conducted using di erent evaluation metrics. The table reports on four di erent algorithm versions (columns 4-7).... In PAGE 13: ...1 in Andrews and Herzberg (1985). The second column in Table3 indicates, on average and for each domain, the location of the nal expression over the coverage plane. Region 1 is the closest to the class-uniform points, region 2 is half-way between the class-uniform points and axis-line A, and nally region 3 is closest to axis-line A (Figure 3b).... In PAGE 13: ... We then take the average over all algorithm versions. Domains in Table3 are ordered based on the values in column 2. We assign a domain to region 1 if the entry on column 2 is within [1:0; 1:5), region 2 if it is within [1:5; 2:5), and region 3 if it is within [2:5; 3:0].... In PAGE 13: ...ersions. Domains in Table 3 are ordered based on the values in column 2. We assign a domain to region 1 if the entry on column 2 is within [1:0; 1:5), region 2 if it is within [1:5; 2:5), and region 3 if it is within [2:5; 3:0]. The third column in Table3 shows the proportion of positive examples on the training set. Each entry is found by averaging the proportion of positive classes on the training set over all runs (all algorithm versions are presented the same training data).... In PAGE 14: ...The last four columns in Table3 estimate, for each algorithm version, the predictive accuracy achieved on each domain by using strati ed 10-fold cross-validation (Kohavi, 1995), averaged over 10 repetitions. Numbers enclosed in parentheses represent stan- dard deviations.... In PAGE 14: ...2 Testing The Utility Of The Distance Measure We now test the utility of the distance-bias measure de ned in equation 9. Our rst experiment uses the last row in Table3 corresponding to the average predictive accuracy for each algorithm over all domains. We compute the absolute di erence in predictive accuracy between each pair of algorithms.... ..."

### Table 4. Classi cation numbers of the real-world problems

2004

"... In PAGE 17: ... But notice that even this data set is not yet a complete practical scenario. The real-world instances are described in Table4 . Notice that two lines, which shall be synchronized to a frequency of T 2 are synchronized explicitly at Table 4.... ..."

Cited by 2

### Table 1. Properties of real-world datasets used in the experiments.

2000

"... In PAGE 4: ...The first set of experiments used 11 datasets from UCI Repository, the collection of datasets provided by the ma- chine learninggroupat the University of California at Irvine (Blake amp; Merz, 1998). The characteristics of these datasets are shown in Table1 . Each of these is a problem whose fea- tures are all either numerical or Boolean (0/1) valued.... ..."

Cited by 2

### Table 1. Properties of real-world datasets used in the experiments.

2000

"... In PAGE 4: ...The first set of experiments used 11 datasets from UCI Repository, the collection of datasets provided by the ma- chine learninggroupat the University of California at Irvine (Blake amp; Merz, 1998). The characteristics of these datasets are shown in Table1 . Each of these is a problem whose fea- tures are all either numerical or Boolean (0/1) valued.... ..."

Cited by 2

### Table 2: Real-World Data Description Soc Kids

"... In PAGE 7: ... Text and meta information were extracted from web pages and the vector space model was applied to represent web pages. Table2 summarizes the two data sets. In order to examine if a predeflned semantics-based hier- archy can provide useful prior knowledge for search, we also compared with the \start from scratch quot; approach: ignore 500 1000 2000 5000 7500 10000 0.... ..."

### Table 2: Real-World Data Description Soc Kids

"... In PAGE 7: ... Text and meta information were extracted from web pages and the vector space model was applied to represent web pages. Table2 summarizes the two data sets. In order to examine if a predefined semantics-based hier- archy can provide useful prior knowledge for search, we also 500 1000 2000 5000 7500 10000 0.... ..."