### Table 4: Parameters for Example 4. (C = 10). Parameter Class 1 Class 2

2002

### Table 1. Linearly increasing number of individuals, concept assertions and role asser- tions for different numbers of universities.

2006

"... In PAGE 12: ... The quadratic shape reveals that this phase should be subject to further optimizations. In Figure 2, average query-answering times for running all 14 LUBM queries on data descriptions for an increasing number of universities are presented (see Table1 for an overview on the number of individuals, concept assertions, and role assertions). We use different modes (A, B, and C) to indicate the effects of optimization techniques.... ..."

Cited by 6

### Table 2 Regions correlating with linearly increasing or decreasing pupil size in other facial expressions

### TABLE 7 Estimated Beta minus Linearly Increasing Theoretical Value: 1970-1996

### TABLE 8 Subperiod Estimated Beta minus Linearly Increasing Theoretical Values Panel A: 1/70-9/79

### Table 21: A summary of the errors for the linear model with the percentile increase for the client in parentheses.

2002

"... In PAGE 69: ... ). In any case, leaving out this attribute increases the error insignificantly: at most by only 44 % in the case of the server and 41 % in the case of the client (Table 16 and Table21 ). Based on these numbers, it may be decided that attribute is not too important in the architectural design of both subsystems.... In PAGE 70: ... probably hides some information about a : when is excluded, the value of the parameter of increases significantly (Appendix 1, section excluded ). 4 a 6 a 3 4 5 a 4 a 3 a 4 a 4 a When excluded from both models in both subsystems, attribute a causes an increase in the error to 367 % in the linear model for the server (Table 11) and to 297 % in the linear model for the client ( Table21 ). The coefficient, as well as the exponent, of the variable that corresponds to attribute gets significant values in both models in both subsystems (Table 27 to Table 31).... In PAGE 70: ... Even though cohesion is present in some models for the client, the value of its coefficient (or exponent) is relatively low compared with those of coupling (attribute ). When leaving cohesion out from the models, the increase in the error is quite insignificant; at most 15 % ( Table21 and Table 26), which means that this architectural attribute is not too important in the architecture of the client. And again, in the case of the server, cohesion has no influence on its architecture.... In PAGE 73: ...8 % to 21.8 %, which is a change of approximately four-fold ( Table21 ). The values of for the server are significantly lower than the corresponding values for the client, especially in the case of the linear model (tables above), but the exclusion of b causes an increase in the error from 2.... ..."

### Table 3: Execution times (minutes) for creating 128 clusters on collections varying in size. Ideal performance is a quadratic increase as the collection size increases and linear decrease as the number of nodes increases

### Table 7, if the number of PNs increases the time for linear approximation increases in some cases. This is due to the increased communication time (resulting from increased number of iterations). Even though the local work on each PN reduces, we observe that the time for communication is much larger than the time for local computation.

1994

"... In PAGE 7: ... Linear Approximation Machine Size Image Size (No. of PNs) 128X128 256X256 512X512 1KX1K 2KX2K 32 30 131 133 407 1721 64 25 131 90 260 905 256 44 128 45 181 297 512 35 151 47 159 234 Table7 : Execution times (in msec) for the linear ap-... ..."

Cited by 1

### Table 1 shows the elapsed time in seconds for the sequential Warshall algo- rithm to complete on a single CPU for matrices ranging in size from 240 x 240 to 1920 x 1920 elements. The times show an expected (linear) increase as the problem size increases.

"... In PAGE 7: ... Table1 . The (sequential) Warshall algorithm Table 2 shows the elapsed time in seconds for the Block and Three-Pass al- gorithms for the same sized matrices as above and varying CPU cluster size.... ..."

### Table 4: Internet prediction assuming linear node increase. We predict the number of edges and e ective diameter of the Internet at the inter-domain level at the beginning of each year.

1999

"... In PAGE 10: ... We can assume that the number of nodes increases a) linearly, or b) by 1:45 each year. The results our shown in Table4 for the linear growth and Table 5 for the 1:45 growth. Given the number of nodes, we calculate the num- ber of edges using Lemma 2 with rank exponent of ?0:81, which is the median of the three observed rank exponents.... ..."

Cited by 780