Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Linear Algebra and Optimisation
Introductory Lecture
N. Anil
BITS Pilani, Hyderabad Campus
anil@hyderabad.bits-pilani.ac.in
July 27, 2024
1 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Scope of the Course
This is an introductory course on
• Linear Algebra
• Optimisation
2 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Introduction to Linear Algebra
What is a linear equation ?
An equation involving one or more variables with constant coefficients
Examples: x = 4, x + y = 1, 2x + y − 3z = 5, etc.
3 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Introduction to Linear Algebra
What is a linear equation ?
An equation involving one or more variables with constant coefficients
Examples: x = 4, x + y = 1, 2x + y − 3z = 5, etc.
A general form :
a1 x1 + a2 x2 + a3 x3 + · · · + an xn = b
3 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Introduction to Linear Algebra
What is a system of linear equations ?
Several linear equations involving the same variables
(
x+y =1
Examples:
x−y =3
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Introduction to Linear Algebra
What is a system of linear equations ?
Several linear equations involving the same variables
( x+ y+ z =1
x+y =1
Examples: , x − y + 3z = 3
x−y =3
2x + 3y − 4z = 5
4 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Introduction to Linear Algebra
System of linear equations : General form
a11 x1 + a12 x2 + a13 x3 + · · · + a1n xn = b1
a21 x1 + a22 x2 + a23 x3 + · · · + a2n xn = b2
.. .. ..
. . .
am1 x1 + am2 x2 + am3 x3 + · · · + amn xn = bm
5 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Types of Linear System of Equations
A given system of equations can be classified into the following three groups:
• Determined system: Number of equations is same as the number of variables.
• Over-determined system: Number of equations is more than the number of
variables.
• Under-determined system: Number of equations is less than the number of
variables.
6 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
System of Equations: Existence and Uniqueness of Solution
Given a system of equations, we have the following possibilities
• Solution exists and it is unique.
7 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
System of Equations: Existence and Uniqueness of Solution
Given a system of equations, we have the following possibilities
(
x+y =1
• Solution exists and it is unique. Example:
x−y =3
• Solution exists but there are infinitely many.
7 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
System of Equations: Existence and Uniqueness of Solution
Given a system of equations, we have the following possibilities
(
x+y =1
• Solution exists and it is unique. Example:
x−y =3
(
x+ y =2
• Solution exists but there are infinitely many.
2x + 2y = 4
• No solution.
7 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
System of Equations: Existence and Uniqueness of Solution
Given a system of equations, we have the following possibilities
(
x+y =1
• Solution exists and it is unique. Example:
x−y =3
(
x+ y =2
• Solution exists but there are infinitely many.
2x + 2y = 4
(
x+y =1
• No solution. Example :
x+y =3
The essence of linear algebra is to investigate the circumstances under which these
scenarios are possible
7 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
System of equations : Matrix form
What is the role of Vectors/Matrices in linear algebra ?
8 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
System of equations : Matrix form
What is the role of Vectors/Matrices in linear algebra ?
(
x+y =1
Consider the linear system:
x−y =3
Let us define " # " # " #
1 1 x 1
A= , x= and b =
1 −1 y 3
The given system can then be written in the compact form as
Ax = b
8 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
System of equations : Matrix form
Another example :
x + y + z = 1
x−y+z =2 (1)
x+y−z =3
Let us define
1 1 1 x 1
A = 1 −1 1 , x = y and b = 2
1 1 −1 z 3
Equation (1) can be written as
Ax = b (2)
The solution of the linear system is nothing but the solution of Ax = b
9 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Solution techniques for Ax = b: Well-known approach
• Solution is given by x = A−1 b
• The solution exists provided A−1 exists
• A−1 = Adj(A)/Det(A)
• A−1 requires the determinant of A
• For large dimensions, computing the determinant is expensive
• Even computer algorithms do not prefer it !
• Need efficient algorithms for the solution of Ax = b !!
10 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Solution techniques for Ax = b: Alternative approaches
• Consider the following linear system
(
x+y =1
x−y =3
• We can write it as " #" # " #
1 1 x 1
=
1 −1 y 3
• Can we get the following through some manipulations (what are they ?)
" #" # " #
1 1 x 1
=
0 −2 y 2
• Now it is easy to find the solution !
• We Transformed the given matrix to a row echelon form. The first part of the
course deals with some of these basic techniques (Gauss elimination, Gauss
Jordan)
11 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Solution techniques for Ax = b: Abstract Concepts
Consider the same invertible system
" #" # " #
1 1 x 1
=
1 −1 y 3
• Replace a column vector in the coefficient matrix with any arbitrary vector
• What happens to the solution. Does it exist ?
• Does that vector has any relation with the existing ones. Is it dependent or
independent (what does this mean ??)
• This leads us to the concept of linear independence/dependence of vectors
• The set of vectors with some special properties is a vector space
12 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Solution techniques for Ax = b: Abstract Concepts
Consider the same invertible system
" #" # " #
1 1 x 1
=
1 −1 y 3
• We generalise: What are those vectors for which the solution exists !
• The concept of vector spaces and related topics precisely deal with it
• What about the eigenvalues and eigenvectors ? Do they say anything about the
nature of the solution or the column vectors ?
• Does the rank of a matrix say anything about the solution ?
• What is the role the coefficient matrix ?
• These abstract concepts give an insight into the nature of the linear system
13 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Linear Algebra: Role of High Performance Computing
• Solving the linear system by hand is limited to 3 or 4 equations
• Beyond this, use of computers is inevitable
• Small number of equations can be solved on a laptop/desktop
• Large number of equations can be solved on a workstation
• Very large number of equations : Cluster
• Many practical applications involve linear systems with billions of equations.
Need Terascale/Petascale/Hexascale Super Computers !!
Need more sophisticated algorithms that are robust, computationally efficient and can
be applied to very large systems with much ease
14 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Linear Algebra: Role of High Performance Computing
Computer implementation of algorithms in linear algebra:
• Traditional codes are based on Fortran, C, C++
• With continuous evolution in computer hardware, porting these codes on modern
HPC platforms is a challenging task
• Code developer needs to tune it frequently
• To circumvent this problem, we can use Julia or Regent
• They are architecture independent, ease code maintenance, code readability
• Regent results in Implicitly parallel solver
• Suits to hybrid parallel codes (CPUs+GPUs)
15 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Linear Algebra: Some of the Application Areas
• Computational biology/chemisty/physics
• Computer graphics, data mining, digital signal processing
• Audio, video and image processing
• Computational Fluid Dynamics
• Computational Structural Mechanics
• Computational Quantum Mechanics
• Computational Finance
• Other related Computational Engineering Science (CES) problems
16 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Numerical Linear Algebra
Numerical implementation of algorithms in linear algebra:
• Numerical algorithms for the solution of linear systems
• To perform matrix operations, inverse, determinant, eigenvalues, etc.
• Standard software library : LAPACK, LAPACK++
• LAPACK uses BLAS (Basic Linear Algebra Subprograms) libraries
• Matlab, Mathematica, Maple use LAPACK libraries
• MATLAB (Matrix Laboratory) is a fourth generation programming language.
Introduced by Cleve Moler from Univ. New Mexico.
• Apps: Wolfram
• PETSc (Portable, Extensible Toolkit for Scientific Computation)
17 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Course Description - Linear Algebra
The linear algebra part has five modules:
• Matrices, System of equations, determinants and inverse of a matrix
• Vector spaces and Linear transformations
• Eigenvalues and eigenvectors
• Numerical Linear Algebra: Gauss elimination and iterative methods for solving
linear systems
• Matrix eigenvalue problems and power method for eigenvalue
18 / 19
Scope of the Course Introduction to Linear Algebra Topics to be covered in Linear Algebra References
Text and Reference books
• Erwin Kreyszig, Advanced Engineering Mathematics, Wiley India, 10th
Edition 2011.
• B. Dubey, Introductory Linear Algebra, Asian Books Pvt Ltd, 2007.
• K Hoffman and R Kunze, Linear Algebra, Pearson Education, 2nd Edition, 2005.
• Elementary linear algebra - S. Andrilli and David Hecker.
• Introduction to linear algebra - G. Strang.
19 / 19