Final05 - L Vandenberghe EE103 Final Exam Solutions Problem 1(20 points The figure shows the function f x =(1 x − 1 1 x − 1 evaluated in IEEE

Info iconThis preview shows pages 1–5. Sign up to view the full content.

View Full Document Right Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: L. Vandenberghe 6/16/05 EE103 Final Exam Solutions Problem 1 (20 points). The figure shows the function f ( x ) = (1 + x ) − 1 1 + ( x − 1) evaluated in IEEE double precision arithmetic in the interval [10 − 16 , 10 − 15 ], using the Matlab command ((1+x)-1)/(1+(x-1)) to evaluate f ( x ). x 1 2 10 − 16 10 − 15 We notice that the computed function is piecewise-constant, instead of a constant 1. 1. What are the endpoints of the intervals on which the computed values are constant? 2. What are the computed values on each interval? Carefully explain your answers. Solution. The first figure shows the rounded values of the numerator and denominator. The second plot shows the result of the division. x (1 + x ) − 1 1 + ( x − 1) ǫ M 2 ǫ M 3 ǫ M 4 ǫ M 5 ǫ M 6 ǫ M 7 ǫ M 8 ǫ M 9 ǫ M 2 ǫ M 4 ǫ M 6 ǫ M 8 ǫ M 10 ǫ M ǫ M 2 ǫ M 3 ǫ M 4 ǫ M 5 ǫ M 6 ǫ M 7 ǫ M 8 ǫ M 9 ǫ M 2 1 2 / 3 4 / 3 1 4 / 5 6 / 5 1 6 / 7 8 / 7 1 8 / 9 2 Problem 2 (20 points). The figure shows m = 20 points ( t i , y i ) as circles. These points are well approximated by a function of the form f ( t ) = αt β e γt . (An example is shown in dashed line.) 1 2 3 4 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 t f ( t ) Explain how you would compute values of the parameters α , β , γ such that αt β i e γt i ≈ y i , i = 1 , . . . , m, (1) using the following two methods. 1. The Gauss-Newton method applied to the nonlinear least-squares problem minimize m summationdisplay i =1 parenleftBig αt β i e γt i − y i parenrightBig 2 with variables α , β , γ . Your description should include a clear statement of the linear least-squares problems you solve at each iteration. You do not have to include a line search. 2. Solving a single linear least-squares problem, obtained by selecting a suitable error function for (1) and/or making a change of variables. Clearly state the least-squares problem, the relation between its variables and the parameters α , β , γ , and the error function you choose to measure the quality of fit in (1). Solution. 1. This problem is a nonlinear least-squares problem minimize g ( x ) = ∑ m i =1 r i ( x ) 2 with variables x = ( α, β, γ ) and r i ( α, β, γ ) = αt β i e γt i − y i . Starting at some initial estimate x (for example, computed with the method of part 2), we repeatedly solve linear least-squares problems minimize bardbl A ( k ) x − b ( k ) bardbl 2 3 with b ( k ) = Ax ( k ) − r ( k ) and A ( k ) = t β 1 e γt 1 α (log t 1 ) t...
View Full Document

This note was uploaded on 03/12/2008 for the course EE 103 taught by Professor L.vandenberghe during the Winter '08 term at Cal Poly Pomona.

Page1 / 9

Final05 - L Vandenberghe EE103 Final Exam Solutions Problem 1(20 points The figure shows the function f x =(1 x − 1 1 x − 1 evaluated in IEEE

This preview shows document pages 1 - 5. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online