Evaluating matrix norms

The definition above of a matrix norm is not directly evaluable in finite time. However, it is possible to determine the value of each of the norms from the elements of a matrix without working through all possible vectors. 

We do this by choosing vectors of unit norm which can be support vectors. In particular,. oo can be evaluated by considering just n vectors, each containing a pattern of Is and -Is matching the pattern of signs in just one row of the matrix. Thus each V oo = 1. 

The corresponding element in AV then has a value equal to the sum of the magnitudes of the entries in that row. The largest of these determines the value of AV and thence A co If Ay denotes the ith entry in the jth row of the matrix A 

The Zqo norm of a matrix is the largest value, taken over the rows of the matrix, of the sum of the absolute values of the entries in that row. 

By taking a support vector with a 1 in the entry corresponding to the maximal column and zeroes elsewhere we can see that 

---

Using matrix property 8, given in the Appendix A (Equation A.20), we can evaluate the norm of the vector r to be r =. 0002. Even though... 

We assume that A is a symmetric and positive semi-definite matrix. The case of interest is when the largest eigenvalue of A is significantly larger than the norm of the derivative of the nonlinear force f. A may be a constant matrix, or else A = A(y) is assumed to be slowly changing along solution trajectories, in which case A will be evaluated at the current averaged position in the numerical schemes below. In the standard Verlet scheme, which yields approximations y to y nAt) via... 

Since this chapter is meant to focus on 4-component type methods I will give some attention to the recent developments that reduce the time spent in evaluation of matrix elements over g. It not trivial to reduce the amount of work associated with the separate upper and lower component basis sets since the norm of the small component wave function may be rather large for heavy atoms. One possibility is to use the locality of the small components of the wave function to replace long-range interactions by a classical interaction [27]. If we distinguish between the upper (large) two and lower (small) two components of the basis 4-spinors... 

The more complex the system or process to be evaluated, the more essential is the need for a HAZOP study. The HAZOP study is conducted in much the same way as the what-if analysis, usually by the same review team. There are minor differences, however, in terminology and approach. In the HAZOP study, certain guidewords are normally used to aid the review team and help identify specific areas where deviations from design intent can occur. Guidewords can include pressure, flow, level, temperature, and power. HAZOP also attempts to identify the severity of the outcome if such deviations from the norm occur as well as the probability or likelihood of occurrence. The hazard risk matrix established and explained in Chapter 2 (Table 2.3) can be used for this purpose since it provides both severity and probability rankings for a given hazardous situation. 

The S l-S o seam of conical intersection in HNCO is analyzed in detail in Chapter 2 of this volume and in Refs. 55 and 56. Here we illustrate the convergence of our search algorithm, Eq. (22), to a point of conical intersection. In this example, is evaluated by a forward difference procedure. can be approximated by a unit matrix if complete energy optimization is not required. However, if the norm of the right hand side, is to be made to vanish rapidly, L 5> 5 is required. Table 2 illustrates typical convergence with L 3- 3 evaluated by forward difference. Here it is important to note that AVi is readily converged to < 0.5 cm . Convergence to approximately this... 

The expansion of the exponential matrix (3.1.2) is rapidly convergent and may often be used for the evaluation of unitary matrices, especially if the anti-Hermitian matrix X has a small norm and high accuracy is not required. An alternative strategy is to diagonalize X ... 

---