### Table 8.7: Variable-sized symbols (available in LATEX)

"... In PAGE 18: ... There are no corresponding size-changing command forms with one argument because such changes are normally only used in the deflnition of commands or in a limited scope. Table8 lists the size-changing commands. declaration size declaration size declaration size {\tiny .... In PAGE 18: ...\Huge ...} size Table8 : Changing the Font Size. EXERCISE 8 Create a LATEX document that formats like the installation script shown in Figure 7.... ..."

### Table 1 Data sets used for regression data set function variable size

"... In PAGE 9: ... Among those data sets, ten are used for regression while the remains are used for classification. The information on the data sets used for regression is tabulated in Table1 . 2-d Mexican Hat and 3-d Mexican Hat have been used by Weston et al.... In PAGE 9: ...nsemble approaches. Plane has been used by Ridgeway et al. [37] in exploring the performance of boosted naive Bayesian regressors. In our experiments, the instances contained in those data sets are generated from the functions listed in Table1 . The constraints on the variables are also shown in Table 1, where U[x, y] means a uniform distribution over the interval determined by x and y.... In PAGE 9: ... The constraints on the variables are also shown in Table 1, where U[x, y] means a uniform distribution over the interval determined by x and y. Note that in our experiments some noise terms have been added to the functions, but we have not shown them in Table1 because the focus of our experiments is on the relative performance instead of the absolute performance of the compared approaches. All the data sets used for classification are from UCI machine learning repository [2], which has been extensively used in testing the performance of diversified kinds of classifiers.... ..."

### Table 1 Data sets used for regression data set function variable size

"... In PAGE 9: ... Among those data sets, ten are used for regression while the remains are used for classification. The information on the data sets used for regression is tabulated in Table1 . 2-d Mexican Hat and 3-d Mexican Hat have been used by Weston et al.... In PAGE 9: ...nsemble approaches. Plane has been used by Ridgeway et al. [37] in exploring the performance of boosted naive Bayesian regressors. In our experiments, the instances contained in those data sets are generated from the functions listed in Table1 . The constraints on the variables are also shown in Table 1, where U[x, y] means a uniform distribution over the interval determined by x and y.... In PAGE 9: ... The constraints on the variables are also shown in Table 1, where U[x, y] means a uniform distribution over the interval determined by x and y. Note that in our experiments some noise terms have been added to the functions, but we have not shown them in Table1 because the focus of our experiments is on the relative performance instead of the absolute performance of the compared approaches. All the data sets used for classification are from UCI machine learning repository [2], which has been extensively used in testing the performance of diversified kinds of classifiers.... ..."

### Table 3.3 OLS Estimates of Expectations Augmented Investment Equation Dependent Variable Ln[I(t)]

### Table 4 Test results with variable-size blocks and SVD techniques for solving FIDAP matrices

in Enhanced

"... In PAGE 17: ...ptimal !. For the variable-size block implementation, we limited the maximum block size to s=200. The maximum number of blocks at each level was also limited to 2000. The test results are given in Table4 . We point out as reference that BILUM with some uniform-size blocks did not converge for several FIDAP matrices tested.... In PAGE 17: ... We point out as reference that BILUM with some uniform-size blocks did not converge for several FIDAP matrices tested. Generally speaking, Table4 shows our best results so far for solving the FIDAP matrices from the point of view of successful convergence. Some undesirable features of the results are the larger... ..."

### Table 4 Test results with variable-size blocks and SVD techniques for solving FIDAP matrices

"... In PAGE 17: ...ptimal !. For the variable-size block implementation, we limited the maximum block size to s=200. The maximum number of blocks at each level was also limited to 2000. The test results are given in Table4 . We point out as reference that BILUM with some uniform-size blocks did not converge for several FIDAP matrices tested.... In PAGE 17: ... We point out as reference that BILUM with some uniform-size blocks did not converge for several FIDAP matrices tested. Generally speaking, Table4 shows our best results so far for solving the FIDAP matrices from the point of view of successful convergence. Some undesirable features of the results are the larger... ..."