### Table 4 Findings for research approach

2002

"... In PAGE 2: ... We further subdivided these categories to reflect a rich set of research approaches. Table4 shows the categories used to classify research approach in this study. The descriptive approach has three subcategories.... In PAGE 5: ...2. Findings for research approach Table4 shows the primary research approaches used by CS researchers. Formulative was by far the dominant research approach representing 79.... ..."

### Table. 4 shows the number of operations per two point input data, the number of cascaded transforms, and the maximum filter length in the decomposed matrices. Theses are important factors to help decide which of the approaches will be used in the practical application. The type- C ORT approach, having a smaller number of operations, is efficient in serial machine application, while the one-length ORT approach has, due to one-tap filters, an implementation facility in parallel machine applications. In particular the one-length ORT approach can be easily implemented in butterfly form, which is very useful in pipeline techniques. For example, a flow- graph of the one-length ORT approach using 4, 4-tap filter bank is shown in Fig. 4. Each butterfly corresponds to its one-length ORT matrix. The inverse can be also implemented reversing the direction of arrows and replacing its inverse matrices.

### Table 3: Different F-ratio criteria applied on the covariance matrices of MFCCs, PCA-transformed MFCCs, SSC and PCA-transformed SSCs.

2003

"... In PAGE 15: ... This F-ratio should be similar to the one calculated using the covariance matrix on the features directly. A simple experiment is performed on this dataset and the results are shown in Table3 . It can be seen that the F-ratios calculated using the determinant of the between- and within-class matrix are affected by small numbers while the F-ratios calculated using the matrix trace are not, i.... ..."

### Table 4-2: Performance of Livermore loops

1988

"... In PAGE 8: ...3 192 15.2 Table4 -1: Performance on Warp array All application programs in the experiment have compile- time loop bounds, and their execution speeds were determined statically by assuming that half of the data dependent branches in conditional statement were taken. All the programs in the sample are homogeneous code, that is, the same cell program is executed by all cells.... In PAGE 8: ...2. Livermore loops The performance of Livermore loops [23] on a single Warp cell is presented in Table4 -2. The Fortran programs were translated manually into the W2 syntax.... ..."

Cited by 460

### Table 3. Head-matching Approach for Direct anaphora

"... In PAGE 4: ... The implemented Head-matching resolution algorithm is: Select all the DDs of the text; For each DD, identify its head-noun; Find a previ- ous occurrence of the same head-noun; If one is found, return the corresponding noun phrase as the antecedent. As seen in Table3 , with this simple approach, 89% of the cases were correctly resolved. Other work consider various extensions of this approach, e.... In PAGE 7: ... Semantic tags approach for Direct anaphors A third experiment, considering the commonest Direct Anaphors yielded the flgures shown in Table 7. Compared with Table3 , semantic tagging for AR was worse than head-matching: unsolved cases amounted to 65 against 9 in the latter approach. This was also due to the lack of semantic tags for proper names and some nouns.... ..."

### Table 1: Example 3 Example 4 This is Example 14 from [5] and Example 2 from [2]. Here, A depends upon a parameter . If ! 0, the system approaches one which is unstabilizable and a conjugate complex pair of the closed loop eigenvalues approaches the imaginary axis. The system matrices are given by

1995

"... In PAGE 14: ... 0 AN?2;N?2 AN?2;N?1 0 0 AN?1;N?1 0 ?1 0 : : : 0 0 ?1 3 7 7 7 7 7 7 7 7 7 7 7 5 where Ak;k = quot; ?1 0 1 0 # ; 1 k N ? 1; Ak+1;k = quot; 0 0 ?1 0 # ; 1 k N ? 2; and E = R = C = In; S = 0; B = diag(1; 0; 1; 0; : : : ; 1; 0; 1); Q = diag(0; 10; 0; 10; : : : ; 0; 10; 0); Stabilizing starting guesses X0 were generated by the method described in [1, 15, 28]. Table1 shows the number of iterations and the Frobenius norm of the last absolute and relative residual. ( ^ X denotes the computed approximation to X .... ..."

Cited by 2

### Table 3. The performance of various approaches where the quality matrices are correct and their variability is lt; 0.1 Our next results shown in Table 4 illustrate the performance of a variety of techniques when less than 10% of discrete columns are perturbed by data quality matrices but the quality matrices have a variance more than 0.1. We find in this situation our approach performs better than before when compared to bagging.

2004

"... In PAGE 3: ... If no result is statistically significant than another, then no entry is in bold. Our first results shown in Table3 illustrate the performance of a variety of techniques when less than 10% of discrete valued columns are perturbed by data quality matrices. We say that the quality matrices have a small variance (less than 0.... In PAGE 4: ....8. We find (shown in Table 5) that the approach performs better than a single model all the time, but is not always better than bagging. As expected the performance improvement is not as great as before ( Table3 and Table 4). Error Single Model Error Bagging (T=910) Error EQPD (T=910) Credit 25.... In PAGE 4: ... The variance of various approaches where the quality matrices are correct and their variability is lt; 0.1 (corresponds with Table3 ). Single Model Bagging (T=1850) EQPD (T=1850) Credit 6.... ..."

Cited by 1

### Table 3: Loop Matrices

1990

Cited by 94

### Table 2: Approaches at each learning loop

"... In PAGE 6: ... While the structural techniques shown examine organisation design from an intra-organisational perspective, a more formalised process within individual organisations and at a network level would have led to a more thorough consideration of the process and structural issues. A possible addition to the design techniques ( Table2 ) is Kumar and Van Dissel [1996]. They examine design from the perspective of inter-organisational relationships.... In PAGE 7: ... However, debate is difficult given their different objectives. Some techniques ( Table2 ) are worthy of further consideration. Revans [1982] sees action learning as important in situations where one is facing conditions previously unknown as managers can help each other by debating when there are no rules.... ..."

### Table 6: two dissimilarity matrices obtained by a monotonic transformation.

"... In PAGE 18: ... An example from De Soete (1983) clearly con rms that additive trees are non meaningful for data measured on a ordinal scale. The two dissimilarity matrices a and b from Table6 are equivalent for the set of monotonic transformations (b is obtained by changing in... ..."