Underdetermined and Overdetermined Linear Algebraic Systems

ES100 March 1, 1999 T.S. Whitten Objectives

● Define underdetermined systems ● Define overdetermined systems ● Examples Review

B 500 N = − ° = ∑ Fx 0; 500 FBC sin 45 0 = ° − = ∑ Fy 0; FBC cos 45 FBA 0 FBC FBA − sin 45° 0 F  − 500 ⋅ BC =  °      1cos42445 4341 1F2BA3  0  coefficients variables Review cont.

− sin 45° 0 F  − 500 ⋅ BC =  °      1cos42445 4341 1F2BA3  0  coefficients variables

The system of matrices above is of the form: Ax = b and can be solved using MATLAB left division thus, x = A\b × results in a 1 2 of values for FBC and FBA Review Summary

● A system of two and two unknowns may yield a unique solution. ● The exception is when the determinant of A is equal to zero. Then the system is said to be singular. ● The left division operator will solve the linear system in one step by combining two matrix operations ●A\B is equivalent to A-1*B Graphical Representation of Unique vs. Singular Systems

unique solution y

singular system

x Underdetermined Systems

● A system of linear equations is may be undetermined if; 1 The determinant of A is equal to zero A = 0 2 The matrix A is not square, i.e. the are more unknowns than there are equations + + = x x 3y 2z 2 1 3 2   2   ⋅ y =   x + y + z = 4 1 1 1 4 z Overdetermined Systems

● The converse of an is an overdetermined system where there are more equations than there are variables ● This situation arises frequently in engineering. For example: suppose a linear relationship is expected between x and y and there are multiple data points. Data Distribution of Linear Phenomena

experimental data y

x Data Distribution of Linear Phenomena

experimental data y y = mx + b

x The line y = mx + b, that best describes this data is obtained by the method of least squares Method of Least Squares

n (x2,y2) =()+−2 J ∑mxibyi i=1 J2 { }J3

}J The line that results in the 1 (x3,y3) minimum value of J is (x ,y ) 1 1 the least squares linear fit to the data. Example

tdata 13.00 25 23.16 20 35.66 48.15 15 511.84 y 613.85 10 data 714.88 5 linear fit 817.67 916.53 0 10 21.65 051015 t The Overdetermined System

● Once the curve fit is obtained, a y-value may be interpolated for any x-value within the x- data range (sometimes extrapolation is possible). ● In the following example, fmins is used to minimize the sum of the squared residuals with respect to the slope and intercept. Flowchart for least.m

load data.txt calculate line assign columns plot line call fmins assign output plot data points and line generate x vector Run least.m Solving The Overdetermined System, Method II

● Solving the overdetermined system is carried out the same way as the other solutions using the left division method m x =   = A \ B b  Method II (cont’d)

● If the system is not overdetermined, the method will not work Overdetermined system cont.

1m + b = 3.00 tdata 13.00 5m + b =11.84 23.16 35.66 10m + b = 21.65 48.15 511.84 613.85 714.88  1 1   3.00  817.67   m   916.53  5 1  ⋅   = 11.84 10 21.65 b  10 1  21.65

Ax B Program least2.m

Start Print slope, intercept Load the data

Generate x, y pairs Build matrix A using fit

Build matrix b Plot data and fit-generated x,y set Left divide (parameters=A\B) Stop Run least2.m