### Table 3 where and b are constants. In the case (i.a) since 6 = 0, by a translation of t, we can set to zero and, by rescaling X with a factor 1= , we can set the parameter

### Table 3. Principle Component Analysis Rescaled Components

"... In PAGE 5: ... The mean and standard deviation of membership tenure indicate that we have a good mix of new and experienced members. We conducted a principal component analysis and the result is shown in Table3 . Five factors (with eigenvalues above 1) are extracted, explaining 68% of total variance.... ..."

### Table 4. Comparison of Precision for NCD with s = 0.5 and CCD*: k = 2 Factors

"... In PAGE 19: ... After re-scaling the CCD by the factor 21/4=1.189, the variances may be compared (see Table4 ). Note that the maximum variance is 0.... ..."

### Table II. Run time comparison (s) of direct and iterative solvers. The direct solver is the multifrontal, supernodal Cholesky factorization in TAUCS; the iterative solver is RMINRES with rescaling and in- complete Cholesky preconditioner and the continuation on the solver tolerance. These timings were obtained on a PC with an AMD Opteron TM252 2.6 GHz 64-bit processor, 8 GB RAM of memory, and the SuSE Linux system.

1798

### Table II. Run time comparison (seconds) of direct and iterative solvers. The direct solver is the multifrontal, supernodal Cholesky factorization in TAUCS; the iterative solver is RMINRES with rescaling and incomplete Cholesky preconditioner and the continuation on the solver tolerance. These timings were obtained on a PC with an AMD Opteron TM252 2.6GHz 64-bit processor, 8GB RAM of memory, and the SuSE Linux system.

2006

### Table 4: Correlation matrix of the logarithms of the bowhead abundance indices. These values were obtained from punt:butt:1999. 1993 can be treated separately from the years 1978-1988, and the BEB likelihood for 1993 can be explicitly included in place of the log-normal distribution assumed for the N4=P4 estimate. In this case, the N4=P4 estimates for 1978-1988 are still assumed to follow a joint distribution with correlation matrix from Table 4 (now excluding the nal row), and they are multiplied by a \bias quot; factor b. The bias factor re-scales the 1978-1988 N4=P4 estimates to re ect the di erence between the 1993 BEB and N4=P4 point estimates. The contribution of the abundance data to the negative log-likelihood function was (excluding constants) therefore:

### Table 3. Homology computation times in seconds for Klein bottle

"... In PAGE 19: ... We apply the same procedure of rescaling as in the two preceding examples. The resulting computation times in seconds are gathered in Table3... In PAGE 25: ...5 sec for rational coefficients. Our cubical representation of Klein bottle presented in the first row of Table3 (for the rescaling factor 1) consists of 1382 four dimensional cubes. This in constrast to the Zomorodian and Carlsson representa- tion, which is two dimensional.... ..."

Cited by 1

### Table 2: Average number of features evaluated per back- ground pattern at a pattern size of 20x20.

2003

"... In PAGE 5: ...oosting algorithm using the basic feature set (i.e., features 1a, 1b, 2a, 2c, and 4a of Figure 2) and stumps as the weak classifiers. As can be seen from Figure 10, Gentle Adaboost outperformed the other two boosting algorithm, despite the fact that it needed on average fewer features (see Table2 , second column). For instance, at a an absolute false alarm rate of 10 on the CMU test set, RAB deteted only 75.... In PAGE 5: ...ate with rescaling factor of 1.2 to 82.7% at a rescaling factor of 1.1. Table2 shows in the second column (nsplit =1) the average number of features needed to be evaluted for background patterns by the different classifiers. As can be seen GAB is not only the best, but also the fastest classifier.... ..."

Cited by 59

### TABLE II. Values of the nonperturbative matrix elements used to calculate c and b pro- duction at the Tevatron; di erent J-values di er simply through a factor (2J+1). The color-octet matrix element for c was determined empirically from CDF data; the color-octet matrix element for b(1P) was found by re-scaling the c matrix element [5]. The color-singlet matrix elements, shown for comparison, were determined from potential model calculations [7]. Matrix Element

### Table 2. Parallelization and rescaling rewriting rules.

in Automatic performance optimization of the discrete Fourier transform on distributed memory computers

2006

Cited by 2