Lec 3

Basic Linear Algebra
Prof. Inder K. Rana

Department of Mathematics
Indian Institute of Technology - Bombay
Lecture – 03
Introduction III
(Refer Slide Time: 00:26)
You can also define addition. So what will be addition? Let us revise addition. Given 2 matrices
𝐴 and 𝐵. If 𝐴 = [𝑎𝑖𝑗 ], 𝐵 = [𝑏𝑖𝑗 ], order is same, 𝑚 × 𝑛. So addition will be 𝑎𝑖𝑗 + 𝑏𝑖𝑗 . So 𝑖𝑗th entry
of the sum is the sum of the 𝑖𝑗th entries of the corresponding matrices. So multiplication of
matrices, for example here, 𝐴 + 𝐵 is same as 𝐵 + 𝐴. This is a nice property like numbers, right.
𝐴 + (𝐵 + 𝐶) is same as (𝐴 + 𝐵) + 𝐶, associative property. Commutative property, associative
property and so on. For example, you can take the matrix to be all entries of 0. So what is 𝐴 + 𝑂
matrix? It is again 𝐴. So 𝑂 is like an identity. So properties similar to that of number addition are
appearing with this kind of addition, right. So we will like to define some multiplication where
some nice properties come out, okay.
Let us just look at the one simple operation called the transpose of a matrix, okay. So what is a
transpose of a matrix? If the matrix is 𝑚 × 𝑛, transpose is just interchange the rows with the
columns. Whatever was the first row, make it the first column of another matrix. So define a new
matrix, first row from where becomes the first column and similarly other are replaced. All the
rows become columns and all columns naturally will become rows, right. So if it was 𝑚 × 𝑛, what
is the new matrix? 𝑛 × 𝑚, right, rows have been.
So here are some simple observations. Transpose is a well-defined matrix, right. We are given all
the entries of it. If A is 𝑚 × 𝑛, obviously 𝐴𝑇 is 𝑛 × 𝑚 and what is transpose of transpose? We are
getting back the matrix. You interchange rows with columns, you again interchange, right, rows
of that with columns, you will get back the matrix. So (𝐴𝑇 )𝑇 = 𝐴, right. So these are elementary
properties of matrices.
So we looked at what is defined as a transpose. So we will call a matrix to be a symmetric if 𝐴 =
𝐴𝑇 . If the interchange of rows and columns does not change the matrix, right and we will call it as
anti-symmetric or normally call skew-symmetric if 𝐴 = −𝐴𝑇 . Does it put any condition on the
matrix 𝐴? If I say a matrix 𝐴 is symmetric, does it put any condition?
Yes, because transpose will change rows to columns. So if 𝐴 was 𝑚 × 𝑛, transpose becomes 𝑛 × 𝑚
and if you have to be equal, that means 𝑚 has to be equal to 𝑛; otherwise, they are not equal from
the equality of matrices. So the concept of symmetric or skew-symmetric matrices is defined only
for square matrices, okay. So that is 1.
So here is now the matrix multiplication. So we have got 2 matrices 𝐴 = [𝑎𝑖𝑗 ] and another matrix
is 𝐵 = [𝑏𝑘𝑙 ]. So here observe 𝐴 is 𝑚 × 𝑛 and B is 𝑛 × 𝑝, okay. Then the product of these 2 matrices
𝐴𝐵, right, is a new matrix, let us call it as 𝐶. So how do I describe 𝐶? So row of the first one,
column of the second, the corresponding entries you multiply and add.
And that naturally puts the condition that the number of columns here should be same as number
of rows in the other. That is why we have put 𝑚 × 𝑛 and 𝑛 × 𝑝, the number of, right. The columns
here is same as the number of rows there. So 𝑚 × 𝑛. So what will be the resulting matrix? The
resulting matrix, this is summed up. So 𝑐𝑗𝑙 , 𝑗 varies 1 to 𝑚, 𝑙 varies from 1 to 𝑝, so that gives you
a matrix of the order is 𝑚 × 𝑛, A; 𝑛 × 𝑝, that is matrix B.

One obvious way of saying that is a nice property is that it is associative. 𝐴(𝐵𝐶) is same as (𝐴𝐵)𝐶.
This way defining of product has got a nice property again, associative, okay. It is not
commutative. 𝐴𝐵 need not be equal to 𝐵𝐴, right. One can give examples for that. So if you want
to define 𝐴𝐵𝐶, you see the condition is put 𝐴 is 𝑚 × 𝑛, 𝐵 is 𝑛 × 𝑝 and 𝐶 is 𝑝 × 𝑞.
This we will not prove it but you can try to write a proof yourself, right. And you can try to verify
it with some simple examples. Product transpose is same as transpose inverted product, right.
(𝐴𝐵)𝑇 is same as 𝐵 𝑇 𝐴𝑇 . One reason that this should be so, because if you take 𝐴𝐵 and transpose.
The rows are going to be interchanged with columns. So similarly if you have 𝐵 𝑇 𝐴𝑇 , already rows
are interchanged with columns, rows are interchanged with columns.
So this is a system. So all the data about the system is captured in these numbers, right. Let us try
to write this slightly more neatly, okay so that we get our system. So let us, this part of the data, I
will call it as a matrix 𝐴. I will call this as b, okay. So what is the order of 𝐴? That is 𝑚 × 𝑛, right.
What is the order of 𝑏? 𝑚 × 1. And what was the variables? 𝑥1 , 𝑥2 , … , 𝑥𝑛 . So let us write them as
a column vector 𝑋. So what is the order of that? 𝑛 × 1, right.
So all the data, the equation can be represented as 𝐴𝑋 = 𝑏. So this is the matrix representation of
a system of linear equation. So we will call the system 𝑚 × 𝑛, right, a system instead of writing
𝑚 × 𝑛 system of 𝑚 linear equations in 𝑛 variables, can be represented as a matrix equation 𝐴𝑋 =
𝑏, where 𝐴 is the matrix which are the coefficients coming in front of the variables, 𝑋 is the
variable, unknown quantity we want to solve and 𝑏 is the vector on the right hand side, right. So
that is the matrix equation. And this is actually equality because when you multiply, right, when
you multiply, you get the first equation, second entry as the second equation, third entry as the
third equation. Equality of matrices give you all the system, right. So the problem is how to solve
this system of linear equations.
And the Gauss elimination method that we had just now seen, it says you can do those 3 operations
that does not change the system of equations, right. Solution for the system. What was our goals?
You can interchange, you can replace the position of the equation. That means what? In this matrix
multiplication, what are you going to do? Changing the order of equation. That means
interchanging the position of one of the rows, interchange of rows, that should not affect the
solution, right.
That does not actually affect the solution. What was the second? You can multiply any row by a
non-0 scalar. That does not change the solution, right. And third you can add any 2 rows if you
like, right. The solution does not change. So you will get a new system which is equivalent to the
earlier one and the solution does not change.
And the idea is that this matrix which is representing, right, it should have lesser number of
coefficients coming, more 0's coming so that you are able to solve the system very easily, right.
So we will see it next time on this Gauss elimination substitution method for a general system of
linear equations, right.

Lec 3

Uploaded by

Copyright:

Available Formats

Lec 3

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lec 3

Uploaded by

Copyright:

Available Formats

Basic Linear Algebra

Prof. Inder K. Rana

(Refer Slide Time: 00:26)

(Refer Slide Time: 07:45)

(Refer Slide Time: 08:54)

(Refer Slide Time: 10:39)

You might also like