The Reflection Method for the

Numerical Solution of Linear Systems

Numerical Methods (MA )
Under the Guidance of Prof(Dr).Nachiketa Mishra
Group :
ME B T VENUMADHAVA REDDY
ME B K SUSMITH
ME B A YASWANTH
ME B G S L MAHA
/
 Table of Contents

▶ Introduction

▶ Problem Definition of the paper

▶ Motivation Of The Paper

▶ Brief Overview Of The Mathematical Methods

▶ Objectives Of This Research

/
 Introduction

• This paper discusses Cimmino’s iterative method for the solution of the system
Ax = b, of n linear equations on Rn , where A is a real n × n sparse matrix that is
initially supposed to be invertible and x, b belong to Rn are column vectors.
• Given two lines on a plane intersecting at Z and a point P ̸= Z, the mirror points of P
with respect to the lines lie on the circle of center Z and radius dist(P, Z).

/ | The Reflection Method for the Numerical Solution of Linear Systems

 The Problem Definition of the Paper
The problem definition presented in the paper focuses on Cimmino's Reflection Method
for solving linear systems numerically.
• Initial Point and Reflections: The method begins with an arbitrary point in Rn and
reflects it across the hyperplanes defined by the linear system.
• Geometrical Approach: It uses geometric reflections to converge to the solution,
which is the centroid of points on a spherical surface centered at the solution point.
The centroid of the reflected points is used as the starting point for the next iteration.
• Convergence and Error Estimates: The paper provides error estimates for
convergence at each iteration and shows that the distance to the solution decreases
with each iteration.
• Applications and Extensions: The reflection method applies to large linear systems
and has been adapted for parallel computing and image reconstruction in
tomography.

/ | The Reflection Method for the Numerical Solution of Linear Systems

 Motivation of the Paper

• The main motivation of the paper is to solve a system of linear equations using
various approaches such as the geometrical approach, numerical approach, and
iterative method.
• The paper demonstrates the interaction of linear algebra in mathematical fields and
how it is widely used in the geometric field as well.
• The study aims to offer a theoretical framework for understanding and analyzing
error bounds associated with Cimmino’s method.

/ | The Reflection Method for the Numerical Solution of Linear Systems

 Brief Overview 2Of The Mathematical Methods
Cimmino's Method for R (n = )

Consider, in the plane R2 , the straight line r of equation

⟨ai , x⟩ = bi , i = 1, 2, ()
(0) (0)
Let’s fix P(0) = P(0) (x(0) ) = P(0) (x1 , x2 ) such that P(0) ∈
/ r.

b − ⟨a, x(0) ⟩
Orthogonal projection of P(0) onto r, R = x(0) + a,
∥a∥2
( )
(0) (0) b − ⟨a, x(0) ⟩
Symmetric point of P w.r.to R, Q = x +2 a.
∥a∥2

/ | The Reflection Method for the Numerical Solution of Linear Systems

 Brief Overview 2Of The Mathematical Methods
Cimmino's Method for R (n = )

Given the initial approximation P(0) = P(0) (x(0) ), then P(1) = P(1) (x(1) ) is the centroid of
unit masses placed at Q(i) , i = 1, 2, of P(0) with respect to the straight line ⟨ai , x⟩ = bi :

X
2
bi − ⟨ai , x(0) ⟩
x(1) = x(0) + ai . ( )
∥ai ∥2
i=1

This will be taken as the starting point of the next iteration, and so on. At step ν + 1,

X
2
bi − ⟨ai , x(ν) ⟩
x(ν+1) = x(ν) + ai . ( )
∥ai ∥2
i=1

/ | The Reflection Method for the Numerical Solution of Linear Systems

 Brief Overview Of The Mathematical Methods
Error bounds
Let’s indicate Z = Z(ξ1 , ξ2 ) the point which solve our system Ax=b. And fix a point
P(0) ̸= Z.Using⟨ai , ξ⟩ = bi , i = 1, 2, we define,

ηi = ⟨ai , x(0) ⟩ − bi = ⟨ai , x(0) − ξ⟩ ( )

Now,
X
2
⟨ai , x(0) − ξ⟩
x (1)
=x (0)
− ai . ( )
||ai ||2
i=1

observe,
2
X
2
ηi2 X
2
1
∥x (1)
− ξ∥ = ∥x
2 (0)
− ξ∥ − 2
2
+ · η i · ai ( )
|ai ∥2 ∥ai ∥2
i=1 i=1

/ | The Reflection Method for the Numerical Solution of Linear Systems

 Brief Overview Of The Mathematical Methods
Error Bounds

With,
X
2 2 X
2  X
2 
pi ηi ai ≤ pi ηi2 · pi ∥ai ∥ ,
2
( )
i=1 i=1 i=1
1
where, pi = ∥ai ∥2
Our Eq ( ) follows as,

∥x(1) − ξ∥ ≤ ∥x(0) − ξ∥ ( )

Now, iterating, we obtain a sequence P(ν) and corresponding estimates

∥x(ν+1) − ξ∥ ≤ ∥x(ν) − ξ∥ ( )

there is equality in Eq ( ) iff x(ν) solves the given system, and hence x(ν) = ξ and also
x(ν+1) = x(ν) for any ν ∈ N.

/ | The Reflection Method for the Numerical Solution of Linear Systems

 Objectives of This Project

• Here we use linear algebra in a well-versed manner for getting the reflections and
orthogonal projections.
• Reflection Algorithm: Introduces Cimmino’s reflection algorithm for solving linear
systems using geometric reflections.
• In this paper, Cimmino’s reflection method is used for solving the system of linear
equations for N= .

/ | The Reflection Method for the Numerical Solution of Linear Systems

 The Reflection Method for the
Numerical Solution of Linear Systems
Thank You

/ | The Reflection Method for the Numerical Solution of Linear Systems