HP -15C Advanced Functions Handbook page 169

Section 4: Using Matrix Operations 1 67
Matrices used: A, B, C, D, and E.
Your subroutine, labeled "E", may use any labels and registers not
listed above, plus the Index register, matrix B, and matrix E
(which should contain your calculated gradient).
To use the program:
1. Enter your subroutine into program memory.
2. Press 11 [T|| DIM iRlTlto reserve registers R
0
through R ±. (Your
subroutine may require additional registers.)
3. Set flag 0 if you're seeking a local minimum; clear flag 0 if
you're seeking a local maximum.
4. Dimension matrix A to n X 1, where n is the number of
variables.
5. Store the required data in memory:
• Store the initial estimate x
0
in matrix A.
• Store a in R
2
.
• Store e in R
3
.
• Store d in R
4
.
• Store N in R
5
.
6. Press |GSB| [A] to view the slopes during the iteration
procedure.
• View the iteration number and the value of/(x).
• If Error 1 appears, press [+] to clear the message. Then
either go back to step 5 and possibly revise parameters as
needed, or press | +| | R/S | to provide one more bounding
search iteration or one more optimization iteration. (If the
descriptor of matrix A was in the display when the error
occurred, the number of bounding search iterations
exceeded N; if the descriptor of matrix C was in the
display, the number of optimization iterations exceeded
N.)
1. Press | R/S | to view the norm of the gradient and to start the
next iteration.
• If the display flashes the norm of the gradient, press f^l
and then recall the values of x in matrix A.