48/50 Linear System Solver
|
02-17-2014, 08:18 AM
Post: #9
|
|||
|
|||
RE: 48/50 Linear System Solver
There are a few issues here..
for me, my flag settings are as follows: -20 Clear (underflow -> 0) -21 Set (Overflow -> error) -22 Clear (Infinite -> error) Then there's another flag that influences the outcome: -54 Use tiny element. With flag -54 Clear (Tiny element ->0), the 'stack solve' (/) returns Infinite Result, and the interactive Linear Solver will return the minimum norm Least Squares solution, because the system is underdetermined. In this mode, the determinant is returned as 0 exactly, because there's a check to see if the matrix contains integers only, and with flag -54 clear, the calculator knows the determinant must be integer and will round the result to the nearest integer. While solving systems, elements less than 1e-14 (relative size) will be set to zero. With flag -54 Set (use tiny element), both methods will return [ -5.41666..e14 -2.0833..e14 -1.25..e14] The determinant now is calculated as -1.2e-13, due to roundoff. So the system is considered NOT singular, and a result can be produced. In any case, it pays to compute the condition number (COND) that returns 9.8e15. As a rule of thumb, the exponent signifies the number of wrong digits in your result, indicating here that the matrix is singular to working precision. Werner In any case 41CV†,42S,48GX,49G,DM42,DM41X,17BII,15CE,DM15L,12C,16CE |
|||
« Next Oldest | Next Newest »
|
User(s) browsing this thread: 1 Guest(s)