You are on page 1of 14

A

SEMINAR ON
THE LEAST SQUARE
SOLUTION OF
LINEAR
INVERSE PROBLEM

UNDER THE GUIIDANCE OF:


DR. DINESH KUMAR
( PROFESSOR)

DELIVERED BY:
REETU
M.TECH (2ND YEAR)
GP-23

Department of Geophysics
Kurukshetra University, Kurukshetra

TABLE OF CONTENTS:

Terminology Used In inverse problem


What Is Inverse Problem
Classification Of Inverse Problem
Linear inverse problem
Least Square Solution of linear
inverse problem
References

Terminology Used In Problem:1. Data points


2. Model parameters
3. Model
1. Data points:- To know the property of the system we
do some
observation these observation or reading
is called data point.

2. Model parameters:- From the data we want to


know some specific properties (or constant) for the
system these constants are known as modal parameter
e.g. In case of earth density,resistivity elastic
parameter is the modal parameter.

Model:-Model is a set of mathematical expression which is used


to relate the data and modal parameter.

What Is Inverse Problem?


It is the problem of determining the nature of a
physical feature by examining the effects it has made.
It is formulated as:
data

model

model parameters

for example:
Potential fields, as in gravity surveying when an
anomaly is interpreted to produce a geological model
parameters.

Classification Of Inverse
Problem:(A) On the based of data point and model
parameter
(B) On the based of model

(A) on the based of data point and model parameter

(1) Over determined inverse problem (data point > model


parameters)
2) underdetermined
inverse problem (data point < model parameters)
(3) Even determined inverse problem (data point = model
parameters)

(B) On the based of model


(1) Linear inverse

problem
(2) Non-linear inverse problem

Linear inverse problem


The simple and best understood linear inverse problem is
that which can be represented in explicit form i.e. we can
separate the data point and model parameter d = Gm
where G is data kernel (may be square and rectangular
matrix) , d is the column vector (matrix) of data points and
m is the column vector of model parameters.
For example : suppose that N temperature measurements T i
are made at depths z i in the earth. The data are then a vector d
of N measurements of temperature , where
d = (T1,T2,T3,T4,----------------TN)T
Let we assume a model T = a + bz, then the intercept a
and slope b are the model parameter of the problem.
m=(a , b)T

According to the model , each temperature observation


must satisfy the
T = a + bz

Solution of linear inverse problem:1. Least square solution


(over determined problem)
2. Minimum length solution
(under-determined problem)
3. Combined function solution
(Mixed determined problem)

Least Square Solution


As we know linear inverse problem can be formulated
as
d=Gm
Where
G is the data kernel
d=(d1,d2,d3,d4,----------------dN)T
m=(m1,m2,m3-----------------mm)T (to be
estimated)
Corresponding to each data point there is
some error which is given by
e=(e1,e2,e3-------------------en)T
d=
Gm+ e

i 1

j 1

i 1

d i Gij m j ei

i 1

i 1

j 1

ei d i Gij m j

So in the least square solution, we minimize the square of the


error by differentiating it with respect to a data point and put
equal to zero I.e.
2

e
i 1

i 1

j1

E (d i G ijm j )
E
mk

m
N
m
j
2(d G m )(0 Gij
)
i
ij j
m
i 1
j1
k

N
m
E
2(d G m )( G ij jk )
mk i 1 i j 1 ij j
If j = k
j k

jk = 1
jk = 0

N
m
E
2(d G m )( G ik )
i
ij j
mk
i 1
j1

E
0
m k

G
i 1 j1

ik

G ij m j G ik d i
i 1

This equation in matrix


form
T
T

G GmG d
T

m (G G) G d
T

-1

References
Menke William, Geophysical data analysis; discrete
inverse theory.
Max A. Meju , Geophysical data analysis:
understanding inverse problem theory and practice

You might also like