Using column scaling, equation (4.12.6), on the problems of...

Using column scaling, equation (4.12.6), on the problems of Exercise 4.16, produces the scaled problems $$ \left(\begin{array}{ll} 1.00 & 2.42 \\ 1.00 & 0.00158 \end{array}\right) \mathbf{x , y}=\left(\begin{array}{l} 3570 \\ 1.47 \end{array}\right),\left(\begin{array}{l} 2.15 \\ 3.59 \end{array}\right) $$ Evaluate $\|\mathbf{r}\|_{\infty}$ for the candidate solutions $\tilde{\mathbf{x}}=\left(\begin{array}{c}-10.0 \\ 1480\end{array}\right)$ and $\tilde{\mathbf{y}}=\left(\begin{array}{c}3.59 \\ -0.595\end{array}\right)$. Could these candidate solutions be considered solutions to problems 'near' the scaled one?

Using column scaling, equation (4.12.6), on the problems of Exercise 4.16, produces the scaled problems
$$
\left(\begin{array}{ll}
1.00 & 2.42 \\
1.00 & 0.00158
\end{array}\right) \mathbf{x , y}=\left(\begin{array}{l}
3570 \\
1.47
\end{array}\right),\left(\begin{array}{l}
2.15 \\
3.59
\end{array}\right)
$$
Evaluate $\|\mathbf{r}\|_{\infty}$ for the candidate solutions $\tilde{\mathbf{x}}=\left(\begin{array}{c}-10.0 \\ 1480\end{array}\right)$ and $\tilde{\mathbf{y}}=\left(\begin{array}{c}3.59 \\ -0.595\end{array}\right)$. Could these candidate solutions be considered solutions to problems 'near' the scaled one?

Key Concepts

Backward Error Analysis

Backward error analysis is a technique used in numerical analysis to assess how much the input data (e.g., the matrix or right-hand side vector) would have to be perturbed so that a given approximate solution becomes an exact solution. This approach helps to determine whether a computed solution can be considered reliable, by assessing if it is the exact solution to a nearby problem, and thus whether any discrepancies in the residual are within acceptable tolerances.

Infinity Norm

The infinity norm of a vector is the maximum magnitude among its components. In error and sensitivity analysis, the infinity norm is often used to measure the worst-case error of a candidate solution, and it provides a straightforward metric for comparing the magnitude of the largest discrepancy in the residual vector.

Column Scaling

Column scaling refers to the process of multiplying the columns of a matrix by appropriate scaling factors to improve the numerical properties of the problem, such as reducing the condition number of the matrix. This technique helps in mitigating numerical errors that can arise from large differences in the magnitude of the matrix entries, thereby stabilizing the computations and making it easier to assess the quality of numerical solutions.

Residual Vector

The residual vector in the context of solving a system of linear equations is defined as the difference between the product of the system matrix and a candidate solution, and the right-hand side vector. It is a measure of how closely the candidate solution satisfies the original system. A small residual norm indicates that the solution is nearly exact, while a larger residual suggests significant error or that the solution is only an approximate one.

Transcript

Please give Ace some feedback