L.Vandenberghe ECE133A(Spring2021) 8. Least squares
L. Vandenberghe
8. Least squares
ECE133A (Fall 2021)
? least squares problem ? solution of a least squares problem ? solving least squares problems
8.1
Least squares problem
given R? and R, find vector R that minimizes
2
- 2 = -
=1 =1
? "least squares" because we minimize a sum of squares of affine functions:
- 2 = ()2,
=1
() = -
=1
? the problem is also called the linear least squares problem
Least squares
8.2
Example
1
2 0
1
=
-1
1
,
=
0
x2
0
0
2
-1
f (x^) + 2 f (x^) + 1
x^
-1-1
0
1
x1
? the least squares solution ^ minimizes
() = - 2 = (21 - 1)2 + (-1 + 2)2 + (22 + 1)2
? to find ^, set derivatives with respect to 1 and 2 equal to zero:
101 - 22 - 4 = 0, -21 + 102 + 4 = 0
solution is (^1, ^2) = (1/3, -1/3)
Least squares
8.3
Least squares and linear equations
minimize - 2
? solution of the least squares problem: any ^ that satisfies ^ - - for all
? ^ = ^ - is the residual vector ? if ^ = 0, then ^ solves the linear equation = ? if ^ 0, then ^ is a least squares approximate solution of the equation ? in most least squares applications, > and = has no solution
Least squares
8.4
Column interpretation
least squares problem in terms of columns 1, 2, . . . , of :
minimize
- 2 =
b
-
2
=1
r = Ax^ - b
Ax^ range(A) = span(a1, . . . , an)
? ^ is the vector in range( ) = span(1, 2, . . . , ) closest to ? geometric intuition suggests that ^ = ^ - is orthogonal to range( )
Least squares
8.5
Example: advertising purchases
? demographic groups; advertising channels ? is # impressions (views) in group per dollar spent on ads in channel ? is amount of advertising purchased in channel ? ( ) is number of impressions in group ? is target number of impressions in group
Example: = 10, = 3, = 1031
Columns of matrix A
Channel 1
2
Channel 2 Channel 3
Target b and least squares result Ax^ 1,500
1,000
Impressions Impressions
1 500
1 2 3 4 5 6 7 8 9 10
Least squares
Group
1 2 3 4 5 6 7 8 9 10
Group
8.6
Example: illumination
? lamps at given positions above an area divided in regions ? is illumination in region if lamp is on with power 1 and other lamps are off ? is power of lamp ? ( ) is illumination level at region ? is target illumination level at region
Example: = 252, = 10; figure shows position and height of each lamp
25m
1 (4.0m)
2 (3.5m)
3 (6.0m)
4 (4.0m)
6 (6.0m)
5 (4.0m)
7 (5.5m)
8 (5.0m)
9 (5.0m) 10 (4.5m)
00
Least squares
25m
8.7
Example: illumination
? left: illumination pattern for equal lamp powers ( = 1) ? right: illumination pattern for least squares solution ^, with = 1
(4.0m)
(3.5m)
(6.0m)
(6.0m) (4.0m)
(4.0m) (5.5m)
(5.0m) (5.0m) (4.5m)
(4.0m)
(3.5m)
1.4
(6.0m) 1.2
(6.0m)
(4.0m) (4.0m)
(5.5m)
1
0.8
(5.0m) (5.0m) (4.5m)
0.6
Number of regions
100
100
50
50
0
0
0
1
2
0
1
2
Intensity
Intensity
Least squares
8.8
................
................
In order to avoid copyright disputes, this page is only a partial summary.
To fulfill the demand for quickly locating and searching documents.
It is intelligent file search solution for home and business.
Related download
- lecture 5 least squares stanford engineering
- 4 qr factorization iit
- chapter 7 least squares estimation
- qr factorization and singular value decomposition
- he ece133a spring2021 8 least squares
- standard format for upper school syllabi
- mathematics program models components
- united nations unece
- mathsteachers
- mm als notes are resources
Related searches
- how to graph least squares regression line
- equation of least squares regression line calculator
- slope of the least squares regression line
- least squares regression line example
- how to use least squares regression
- least squares equation
- the least squares regression line calculator
- method of least squares equation
- least squares regression calculator
- least squares regression equation calculator
- weighted least squares regression excel
- least squares regression method