Numerical Methods in Computational Finance. Daniel J. Duffy

Чтение книги онлайн.

Читать онлайн книгу Numerical Methods in Computational Finance - Daniel J. Duffy страница 53

Numerical Methods in Computational Finance - Daniel J. Duffy

Скачать книгу

if upper A minus upper B is positive. These kinds of matrices have many applications, for example Markov chains and stochastic matrix theory. They are also relevant when we construct monotone finite difference schemes and M-matrices to approximate the solution of differential equations, as we shall see later in this book.

      5.6.6 Irreducible Matrices

      A matrix A is said to be reducible if there exists a permutation matrix P such that:

italic upper P upper A upper P Superscript down-tack Baseline equals Start 2 By 2 Matrix 1st Row 1st Column upper A 11 2nd Column upper A 12 2nd Row 1st Column 0 2nd Column upper A 22 EndMatrix comma left-parenthesis upper A 11 comma upper A 12 comma upper A 22 a r e square matrices right-parenthesis period

      The matrix A is called irreducible if no such permutation matrix exists.

      The matrix A is said to be diagonally dominant if:

      (5.21)StartAbsoluteValue a Subscript italic j j Baseline EndAbsoluteValue greater-than-or-equal-to sigma-summation Underscript StartLayout 1st Row k equals 1 2nd Row k not-equals j EndLayout Overscript n Endscripts StartAbsoluteValue a Subscript italic j k Baseline EndAbsoluteValue identical-to rho Subscript j Baseline for-all j equals 1 comma ellipsis comma n where upper A equals left-parenthesis a Subscript italic i j Baseline right-parenthesis comma 1 less-than-or-equal-to i comma j less-than-or-equal-to n period

      Some results that are used in PDE applications are:

      1 If A is strictly diagonally dominant, then it is invertible.

      2 If A is irreducible and diagonally dominant and if for at least one j, then A is invertible.

      5.6.7 Other Kinds of Matrices

      A Metzler matrix A satisfies a Subscript italic i j Baseline greater-than-or-equal-to 0 comma i not-equals j where upper A equals left-parenthesis a Subscript italic i j Baseline right-parenthesis period

      A Z-matrix is a negated Metzler matrix a Subscript italic i j Baseline less-than-or-equal-to 0 comma i not-equals j where upper A equals left-parenthesis a Subscript italic i j Baseline right-parenthesis.

      An M-matrix is a Z-matrix with eigenvalues whose real parts are non-negative. An M-matrix can be expressed in the form upper A equals italic s upper I minus upper B comma upper B equals left-parenthesis b Subscript italic i j Baseline right-parenthesis comma b Subscript italic i j Baseline greater-than-or-equal-to 0 comma i comma j equals 1 comma ellipsis comma n. The scalar s is at least as large as the maximum of the moduli of the eigenvalues of B, and I is the identity matrix.

      Some applications of M-matrices are from mathematics and economics, for example:

       Establish bounds on eigenvalues.

       Convergence criteria for iterative methods.

       Discretisations of PDEs (for example, in combination with exponential fitting) and monotone finite difference schemes.

       Finite Markov chains.

       Population dynamics.

      Finally, L-matrices are defined by: upper L equals left-parenthesis l Subscript italic i j Baseline right-parenthesis comma l Subscript italic i i Baseline greater-than 0 comma l Subscript italic i j Baseline less-than 0 comma i not-equals j period

      The Cayley Transform refers to a group of related concepts. The relevance to our work is that it appears when proving the stability of finite difference schemes for two-factor option pricing problems as we shall discuss in Chapters 18, 22 and 23. It has been applied in fixed-income pricing as discussed in Davidson and Levin (2014).

      In general, we define:

      (5.22)upper Q equals left-parenthesis upper I minus sigma upper A right-parenthesis left-parenthesis upper I plus sigma upper A right-parenthesis Superscript negative 1 Baseline comma sigma greater-than 0 period

      In the case where A is positive definite, we can compute the norm of Q as follows:

StartLayout 1st Row 1st Column double-vertical-bar upper Q double-vertical-bar squared 2nd Column equals max Underscript phi Endscripts StartFraction left-parenthesis upper Q phi comma upper Q phi right-parenthesis Over left-parenthesis phi comma phi right-parenthesis EndFraction equals max Underscript phi Endscripts StartFraction left-parenthesis left-parenthesis upper I minus sigma upper A right-parenthesis phi comma left-parenthesis upper I minus sigma upper A right-parenthesis phi right-parenthesis Over left-parenthesis left-parenthesis upper I plus sigma upper A right-parenthesis phi comma left-parenthesis upper I plus sigma upper A right-parenthesis phi right-parenthesis EndFraction comma 2nd Row 1st Column Blank 2nd Column equals max Underscript phi Endscripts StartFraction left-parenthesis phi comma phi right-parenthesis minus 2 sigma left-parenthesis upper A phi comma phi right-parenthesis plus sigma squared left-parenthesis upper A phi comma upper A phi right-parenthesis Over left-parenthesis phi comma phi right-parenthesis plus 2 sigma left-parenthesis upper A phi comma phi right-parenthesis plus sigma squared left-parenthesis upper A phi comma upper A phi right-parenthesis EndFraction less-than-or-equal-to 1 period EndLayout

      This result was proved in Kellogg (1964), and it is an important lemma to prove unconditional stability of splitting schemes. In the same way it is possible to show that double-vertical-bar left-parenthesis upper I plus sigma upper A right-parenthesis Superscript negative 1 Baseline double-vertical-bar less-than 1.

       Appendix : The Schrödinger Equation

      In

Скачать книгу