Main content

## Linear algebra

### Course: Linear algebra > Unit 1

Lesson 7: Null space and column space- Matrix vector products
- Introduction to the null space of a matrix
- Null space 2: Calculating the null space of a matrix
- Null space 3: Relation to linear independence
- Column space of a matrix
- Null space and column space basis
- Visualizing a column space as a plane in R3
- Proof: Any subspace basis has same number of elements
- Dimension of the null space or nullity
- Dimension of the column space or rank
- Showing relation between basis cols and pivot cols
- Showing that the candidate basis does span C(A)

© 2023 Khan AcademyTerms of usePrivacy PolicyCookie Notice

# Column space of a matrix

Introduction to the column space of a matrix. Created by Sal Khan.

## Want to join the conversation?

- why is it called column space when it just the linear combination of vectors,or am i missing something?(6 votes)
- Column space is not as clumsy as "linear combination of vectors". Eventually you'll talk in terms of column space, null space, etc. The word "space" as in "vector space" has it's own meaning and properties so having the word space in the help to remind you it has those properties. I don't think you're not missing anything (unless I am too :))(15 votes)

- HI! Why do you always define the matrix as a bunch of column vectors? Could you also say that elements of each row are components of a row vector or transposed column vector?

Thank you!(7 votes)- Yes, you could also work out all of linear algebra using row vectors instead of column vectors (but you would have to adapt most of the operations for this). Historically mathematicians have always used column vectors, so we continue to use them that way.

There is no other reason than historical custom to continue doing so. And since everyone in the world will do the same, it's good practice to adapt to the standard.(8 votes)

- Why have two names for the same property of a matrix? It seems to me that the span of the column vectors of the matrix is just a second name for the column space of the matrix in all cases. Are there subspaces for which this is not true?(4 votes)
- There are many things that have multiple names that represent the exact same thing, ("antiderivative" and "indefinite integral" comes to mind) it might be because they were discovered/invented by different people or following different methodologies, and it's simpler to just retain all the names than try to unify everything.

You are right, though: "matrix column space" and "span of the column vectors of a matrix", will always represent the exact same thing.(6 votes)

- Is a Column Space the exactly same thing as the Span?(4 votes)
- No. It's a span, of the column vectors. You can take the span of any set of vectors.(3 votes)

- I'm trying to tell the difference between a column space and a span. This is what I understand so far, can someone tell me if it's correct:

The column space is all the possible vectors you can create by taking linear combinations of the given matrix.

In the same way that a linear equation is not the same as a line, a column space is similar to the span, but not the same. The column space is the matrix version of a span. The Span is the graphical representation of the column space.(3 votes)- You take the span of a set of vectors. You take the column space of a matrix. The column space of a matrix is the span of its column vectors. Taking the span of a set of vectors returns a subspace of the same vector space containing those vectors.(2 votes)

- sorry, but my question is more of a general question. I understand all the previous concepts in your videos very well but I feel I need to solve some exercises on my own, can you recommend any exercises or books for me, please ?(3 votes)
- Linear Algebra and Its Applications by David C. Lay has some pretty good exercises.(1 vote)

- I don't see how you can multiply x1 times v1 if v1 is a column vector with m elements and x1 is a column vector with n elements. How is that possible?(2 votes)
- You aren't. x1 is not a column vector. x1 is the first element of the column vector x. In other words, x1 is a scalar.(2 votes)

- So is Metrix A(V1,V2,...Vn) a valid subpace?(2 votes)
- Hi, I really like these videos and the way you are explaining. I have a small question in this context. Can we show a given matrix is lie in the column space of another matrix?

Thanks(2 votes) - One thing I am not getting clarity here. Sal uses the valid subspace conditions as an assertion to the set of vectors.

A valid subspace (V) requires S.<a> = Sc1.<v1> + Sc2.<v2> + .. Scn.<vn> where all <> are my vectors to be within the (V). It should also satisfy the following conditions

1. <0> should be within it

2. closed under <V> any scalar multiplication

3. closed under <V> addition of any random two vectors within <V>

Span (v1,v2,...vn) <- R(n).. this span covers the entire R(n). it is more like the subspace of a room is the room itself and it is a valid subspace.

Question - do we need to prove all the time with arbitrary conditions and then assert it belongs to the span?(2 votes)

## Video transcript

We spent a good deal of time on
the idea of a null space. What I'm going to do in this
video is introduce you to a new type of space that can be
defined around a matrix, it's called a column space. And you could probably guess
what it means just based on what it's called. But let's say I have
some matrix A. Let's say it's an
m by n matrix. So I can write my matrix A and
we've seen this multiple times, I can write it as a
collection of columns vectors. So this first one, second one,
and I'll have n of them. How do I know that
I have n of them? Because I have n columns. And each of these column
vectors, we're going to have how many components? So v1, v2, all the way to vn. This matrix has m rows. So each of these guys are going
to have m components. So they're all members of Rm. So the column space is defined
as all of the possible linear combinations of these
columns vectors. So the column space of A, this
is my matrix A, the column space of that is all the linear
combinations of these column vectors. What's all of the linear combinations of a set of vectors? It's the span of
those vectors. So it's the span of vector
1, vector 2, all the way to vector n. And we've done it before
when we first talked about span and subspaces. But it's pretty easy to show
that the span of any set of vectors is a legitimate
subspace. It definitely contains
the 0 vector. If you multiply all of these
guys by 0, which is a valid linear combination added up,
you'll see that it contains the 0 vector. If, let's say that I have some
vector a that is a member of the column space of a. That means it can be represented
as some linear combination. So a is equal to c1 times vector
1, plus c2 times vector 2, all the way to Cn
times vector n. Now, the question is, is this
closed under multiplication? If I multiply a times some new--
let me say I multiply it times some scale or s, I'm just
picking a random letter-- so s times a, is this
in my span? Well s times a would be equal
to s c1 v1 plus s c2 v2, all the way to s Cn Vn Which is
once again just a linear combination of these
column vectors. So this Sa, would clearly
be a member of the column space of a. And then finally, to make sure
it's a valid subspace-- and this actually doesn't apply just
to column space, so this applies to any span. This is actually a review of
what we've done the past. We just have to make sure it's
closed under addition. So let's say a is a member
of our column space. Let's say b is also a member
of our column space, or our span of all these
column vectors. Then b could be written as b1
times v1, plus b2 times v2, all the way to Bn times Vn. And my question is, is a plus b
a member of our span, of our column space, the span
of these vectors? Well sure, what's a plus b? a
plus b is equal to c1 plus b1 times v1, plus c2 plus
v2 times v2. I'm just literally adding
this term to that term, to get that term. This term to this term
to get this term. And then it goes all the way
to Bn and plus Cn times Vn. Which is clearly just
another linear combination of these guys. So this guy is definitely
within the span. It doesn't have to be
unique to a matrix. A matrix is just really just
a way of writing a set of column vectors. So this applies to any span. So this is clearly
a valid subspace. So the column space of a is
clearly a valid subspace. Let's think about other ways we
can interpret this notion of a column space. Let's think about it in terms of
the expression-- let me get a good color-- if I were to
multiply my-- let's think about this. Let's think about the set of all
the values of if I take my m by n matrix a and I multiply
it by any vector x, where x is a member of-- remember x has
to be a member of Rn. It has to have n components in
order for this multiplication to be well defined. So x has to be a member of Rn. Let's think about
what this means. This says, look, I can take any
member, any n component vector and multiply it by a,
and I care about all of the possible products that this
could equal, all the possible values of Ax, when I can
pick and choose any possible x from Rn. Let's think about
what that means. If I write a like that, and if
I write x like this-- let me write it a little bit better,
let me write x like this-- x1, x2, all the way to Xn. What is Ax? Well Ax could be rewritten as
x1-- and we've seen this before-- Ax is equal to x1 times
v1 plus x2 times v2, all the way to plus Xn times Vn. We've seen this multiple
times. This comes out of our
definition of matrix vector products. Now if Ax is equal to this,
and I'm essentially saying that I can pick any vector x in
Rn, I'm saying that I can pick all possible values of the
entries here, all possible real values and all possible
combinations of them. So what is this equal to? What is the set of
all possible? So I could rewrite this
statement here as the set of all possible x1 v1 plus x2 v2
all the way to Xn Vn, where x1, x2, all the way to Xn, are
a member of the real numbers. That's all I'm saying here. This statement is the
equivalent of this. When I say that the vector x can
be any member of Rn, I'm saying that its components
can be any members of the real numbers. So if I just take the set of all
of the, essentially, the combinations of these column
vectors where their real numbers, where their
coefficients, are members of the real numbers. What am I doing? This is all the possible linear
combinations of the column vectors of a. So this is equal to the span
v1 v2, all the way to Vn, which is the exact same thing
as the column space of A. So the column space of A, you
could say what are all of the possible vectors, or the set of
all vectors I can create by taking linear combinations of
these guys, or the span of these guys. Or you can view it as, what are
all of the possible values that Ax can take on if
x is a member of Rn? So let's think about
it this way. Let's say that I were to tell
you that I need to solve the equation Ax is equal to-- well
the convention is to write a b there-- but let me
put a special b there, let me put b1. Let's say that I need to
solve this equation Ax is equal to b1. And then I were to tell you--
let's say that I were to figure out the column space of
A-- and I say b1 is not a member of the column space of
A So what does that tell me? That tells me that this right
here can never take on the value b1 because all of the
values that this can take on is the column space of A. So if b1 is not in this, it
means that this cannot take on the value of b1. So this would imply that this
equation we're trying to set up, Ax is equal to b1,
has no solution. If it had a solution, so let's
say that Ax equals b2 has at least one solution. What does this mean? Well, that means that this, for
a particular x or maybe for many different x's,
you can definitely achieve this value. For there are some x's that when
you multiply it by a, you definitely are able
to get this value. So this implies that b2 is
definitely a member of the column space of A. Some of this stuff on some level
it's almost obvious. This comes out of
the definition of the column space. The column space is all of the
linear combinations of the column vectors, which another
interpretation is all of the values that Ax can take on. So if I try to set Ax to some
value that it can't take on, clearly I'm not going to
have some solution. If I am able to find a solution,
I am able to find some x value where Ax is equal
to b2, then b2 definitely is one of the values that
Ax can take on. Anyway, I think I'll
leave you there. Now that you have at least a
kind of abstract understanding of what a column space is. In the next couple of videos
I'm going to try to bring everything together of what we
know about column spaces, and null spaces, and whatever else
to kind of understand a matrix and a matrix vector product from
every possible direction.