### Table 5. Memory and learning test scores by PCB exposure quartile. WMS logical memory, delayed recall CVLT Semantic Cluster Ratio CVLT List A, Trial 1

"... In PAGE 4: ... Lead and mercury exposure did not significantly impact verbal delayed recall and were not retained in the final model. To further illustrate the negative impact of PCB exposure on verbal delayed recall, we divided the sample into quartiles based on PCB exposure and calculated observed and predicted recall scores ( Table5 ). All but 5 of the 45 subjects in the highest PCB exposure quartile were fish eaters.... In PAGE 4: ...nd List A, Trial 1 (p = 0.037). DDE, lead, and mercury exposure were not related to these outcomes. The observed and predicted scores for the Semantic Cluster Ratio and List A, Trial 1, by PCB exposure quartile are shown in Table5 . For both tests, we observed a clear downward trend in scores with increasing PCB exposure.... ..."

### Table 2: Estimated Total Delay of Paths between Each PE and a Shared Memory

2003

"... In PAGE 103: ... Table2 shows estimated delays for the GGBA estimated layout shown in Fig- ure 30. The second column shows estimated interconnect delays described in Sec- tion 6.... In PAGE 103: ... Table 3 shows the number of clock delay cycles that will be inserted into a memory cycle for the cases that a GGBA system has three different bus clocks, respectively. The total delays shown in Table2 are divided by each bus clock period in order to obtain the number of clock delays shown in Table 3. Figure 32 describes the sequence of MBI module generation, which is a module generation procedure of our bus synthesis tool (BusSynth) that will be described in Section 6.... In PAGE 125: ...3.1, and the delay values are shown in Table2 . Finally, the third system, GGBA III, has a memory controller that operates with a maximum estimated delay on all connections between the PEs and the shared memory.... ..."

Cited by 15

### Table 1: Comparisons of serial execution time for direct, CG, and ILUCG linear system solvers when used for the transient simulation of the circuit in Figure 1, where gf = 3.0e- 5 and g, has a conductance of le - 3 when linearized about zero.

1990

"... In PAGE 13: ... When simulating grid-based signal processors, where the coupling between subcircuits is restricted to nonlinear resistors, the Newton iteration equation will be such that its solution can be efficiently computed by iterative algorithms like conjugate-gradient squared (CGS) [11, 1]. To demonstrate this, in Table1 , we compare the CPU time required to compute the transient analysis of the network in Figure 1 using several different matrix solution algorithms to solve the Newton iteration equation. This problem is hard for an iterative method because, though not described here, the transient analysis is performing a continuation on the nonlinear resistor elements that changes the conditioning of the matrix with time (see [5] for details).... ..."

### Table 7: Effects of different features on V3.0

in Effective adaptation of a hidden Markov model-based named entity recognizer for biomedical domain

2003

"... In PAGE 7: ... In order to evaluate the contributions of differ- ent features, we evaluate our system using different combinations of features (Table 7). From Table7 , several findings are concluded: 1) With only Fsd, our system achieves a basic level F-measure of 29.4.... ..."

Cited by 17

### Table 7: Effects of different features on V3.0

in Effective adaptation of a hidden Markov model-based named entity recognizer for biomedical domain

2003

"... In PAGE 7: ... In order to evaluate the contributions of differ- ent features, we evaluate our system using different combinations of features (Table 7). From Table7 , several findings are concluded: 1) With only Fsd, our system achieves a basic level F-measure of 29.4.... ..."

Cited by 17

### Table 1: ARLs, alarm and equilibrium rates of RLu( 0), for u = 0; 3, and 0 = 0; 0:02.

"... In PAGE 8: ... Example 6 | Consider the upper one-sided CUSUM scheme for the detection of upward shifts in the expected number of defectives per random sample of size n that makes use of the summary statistic de ned by Equation (5) with: reference value k = 3; upper control limit x = 6; initial values u = 0 (scheme with no head start) and u = 3 (scheme with a 50% head start); and XN i:i:d: Binomial(n = 100; p0 = 0:02) in the absence of assignable causes. In this case, the in-control run length, RLu = RLu(0), has a discrete phase-type distribution with in-control sub-stochastic matrix equal to Q = Q(0) = 2 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4 0:8590 0:0902 0:0353 0:0114 0:0031 0:0007 0:0002 0:6767 0:1823 0:0902 0:0353 0:0114 0:0031 0:0007 0:4033 0:2734 0:1823 0:0902 0:0353 0:0114 0:0031 0:1326 0:2707 0:2734 0:1823 0:0902 0:0353 0:0114 0 0:1326 0:2707 0:2734 0:1823 0:0902 0:0353 0 0 0:1326 0:2707 0:2734 0:1823 0:0902 0 0 0 0:1326 0:2707 0:2734 0:1823 3 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 5 : (21) The parameters yield to a scheme with ARLs at the in-control value np0 = 2 equal to ARLu(0) = 1015:71; 995:07, for u = 0; 3 (respectively), as shown by Table1 , where we can also nd out-of-control ARL values for an upward shift with magnitude = p p0 = 0:02, and some values of the alarm rate and the equilibrium rate functions of RLu( 0); u = 0; 3 and 0 = 0; 0:02. The use of a 50% head start yields a mild relative reduction (of 2.... In PAGE 9: ...The alarm rate values also show how unlikely (likely) is the emission of a false alarm (of a correct signal) at sample m, given that no previous signal has been triggered. With no head start the alarm rate of the control scheme increases as we proceed with the sampling procedure (in opposition with the constant alarm rate of the geometric run length of Shewhart schemes), as shown by Columns 2 and 4 of Table1 and by Figure 1. Thus, signalling in the absence of assignable causes, given that no observation has previously exceeded the upper control limit, becomes more likely, as we proceed with the sampling procedure.... In PAGE 10: ... We should also add that the limiting form of the probability function of the RL is geometric-like with parameter 1 ( ), where ( ) is the maximum real eigenvalue of Q( ) (see Brook and Evans (12)), regardless of the initial value u of the summary statistic. Thus, it comes as no surprise that the values of the alarm rate functions of RL0(0) and RL3(0), and RL0(0:02) and RL3(0:02) rapidly converge to lim m!+1 RLu(0)(m) = 1 (0) = 0:000987 (22) lim m!+1 RLu(0:02)(m) = 1 (0:02) = 0:223611; (23) respectively, as seen in Table1 and Figure 1. A similar convergence hold for the equilibrium rate function: lim m!+1 rRLu( 0)(m) = 1= ( 0); 0 = 0; 0:02: (24) Finally, if we compare Columns 2i and 2i + 1, i = 1; 2; 3; 4, of Table 1 we can see that... In PAGE 10: ... Thus, it comes as no surprise that the values of the alarm rate functions of RL0(0) and RL3(0), and RL0(0:02) and RL3(0:02) rapidly converge to lim m!+1 RLu(0)(m) = 1 (0) = 0:000987 (22) lim m!+1 RLu(0:02)(m) = 1 (0:02) = 0:223611; (23) respectively, as seen in Table 1 and Figure 1. A similar convergence hold for the equilibrium rate function: lim m!+1 rRLu( 0)(m) = 1= ( 0); 0 = 0; 0:02: (24) Finally, if we compare Columns 2i and 2i + 1, i = 1; 2; 3; 4, of Table1 we can see that... ..."

### Table 1. The energy of a multiplet have the expression Efmig = M X

"... In PAGE 9: ... (z) = QM i=1 z?L=2 i (z) : Hj i = X n1;:::;nM(?1)P ni HCS ? ML2 4 (!n1? ; :::; !nM? ) ? n1; :::; ? nM j i (4:10) The values of the coupling constants c1;2 in HCS are the values given in Table1 and the constant is equal to 2. Note that, for these values, the energy in (4.... ..."

### Table 14. Following the Boundary: r = 3:0, Parameter ratio (x:50:30)

1996

"... In PAGE 32: ... The results are illustrated in Figure 11, where each plotted point is the median of ten runs. Various parameters to IG were tried, and a sample of typical results for r = 3:0 and r = 3:5 can be found in Table14 and Table 15. We see that as r is increased (decreasing k) the cost of coloring is reduced sharply.... In PAGE 32: ... Perhaps not evident from the graph is the fact that even for r = 3:0 the number of iterations is approximately doubling for each increase of 1000 vertices and for r = 3:5 it doubles about every 1500 vertices. See Table14 and Table 15. Thus, we have not escaped the exponential growth inherent in Ku cera apos;s result [14].... ..."

Cited by 29

### Table 3: #0Cxed portion of a system description vector

1994

"... In PAGE 9: ... To determine the basis for the #0Cxed portion of the description vector, we reviewed 18 algorithms from the existing scheduling literature #5B2#7B5, 7, 13, 17#7B20, 22#7B24, 26#7B28, 30, 31#5D. Table3 depicts the resulting data set. We found that only twocharacteristics|processor speed and inter-processor communication time estimates|were used by more than four algorithms.... ..."

Cited by 22