If you're seeing this message, it means we're having trouble loading external resources on our website.

If you're behind a web filter, please make sure that the domains *.kastatic.org and *.kasandbox.org are unblocked.

Main content

Matrices as transformations of the plane

We can think of a 2X2 matrix as describing a special kind of transformation of the plane (called "linear transformation"). By telling us where the vectors [1,0] and [0,1] are mapped to, we can figure out where any other vector is mapped to. Created by Sal Khan.

Want to join the conversation?

• Would be good to have more practice to break up this string of videos.
(21 votes)
• More practice always makes perfect.
(1 vote)
• Can 3x3 matrices be used for 3D transformations?
(11 votes)
• Exactly correct! For any n x n matrix for any number n, that transformation can be applied to the nth dimension. Not sure how a 4x4 matrix can be visually applied; however, mathematically speaking it works!
(9 votes)
• What's a weighted sum?
(5 votes)
• SUm where each say component is multiplied by some constants( treated as weights ) as in Linear combination
(8 votes)
• this thing just went above my head don't even know what's a transformation and what was Sal trying to do in this video.
(6 votes)
• I'm wondering if this just happened by occasion that the matrices were applied to Cartesian coordinates for transformations and were found useful for that or this is what they were invented for originally?
(4 votes)
• I'm not entirely sure, but it happens a lot in maths that things are created without external subjects in mind, and that those external subjects are intentionally applied to fit with these systems so as not to create any unnecessary definitions. If there is no application that can be made, it is usually left as undefined, an example being addition of matrices with different dimensions, so to specifically answer your question, I think that matrices were applied to Cartesian coordinates as there were deemed as a useful extension.
(4 votes)
• Does vector have something to do with matrices?
(2 votes)
• as far as i know a vector is a type of matrix which has only one entry in a row or column while more than one in a column and row, respectively (in other words, a slim matrix)

but if you mean a vector in a sense of physics which has some properties like magnitude and direction, i'm not sure either
(3 votes)
• for the matrix that is called the identity transformation, why does the point map back to it self. Shouldn't it become (1,1) instead of still in the origin?
(2 votes)
• Any transformation that you can represent with a matrix will map the origin to itself. Multiplying any matrix by a zero matrix (that is, applying the transformation to the origin) gives you back the zero matrix.
(2 votes)
• Wait, so the 2x2 matrix transformations are based on their current co-ords. For example, does that mean that a [2 1] [1 2]
transformation on a point at (2, 3) move it to (7, 8), but that same transformation on (0,0) moves it to (0,0)? How does that even work?
(2 votes)
• The point [0, 0] is always fixed. You don't always have to change the point to have a transformation.
(1 vote)
• A `n x 1` vector is transformed if it's multiplied by a `n x n` matrix.

So what does it mean if `n x 1` is multiplied by a non-square matrix?
`3 x 2` * `2 x 1` for example?
(1 vote)
• As you see in your example, the middle terms MUST be the same in order to multiply matrices, the resulting matrix will be the size of the two outside numbers, so you would end up with a 3x1 matrix. If you tried to reverse positions and did 2x1*3x2, the middle numbers are not the same, so they cannot be multiplied. Is this enough to answer your question?
(2 votes)
• In this video, the coordinate of a point is represented a 2 by 1 matrix rather than a 1 by 2 matrix. Regarding the multiplication rule in linear algebra, only in this form (2 by 1 in the video), linear transformation can be explained by a weighted sum? Simply put, if a point's coordinate is in the form of 1 by 2 matrix, it is no longer a weighted sum. It took me some time to figure out what's going on here.
(1 vote)
• Yes, you are correct. In linear algebra, a linear transformation can be represented as a matrix-vector multiplication, where the matrix represents the transformation and the vector represents the coordinates of the point being transformed.

When representing a point's coordinates as a 2 by 1 matrix (i.e., a column vector), the matrix-vector multiplication can be interpreted as a weighted sum of the components of the vector, where the weights are the entries of the matrix. This is because each entry in the resulting vector is obtained by taking the dot product of a row of the matrix with the column vector, which is equivalent to a weighted sum of the components of the vector.

On the other hand, if the point's coordinates are represented as a 1 by 2 matrix (i.e., a row vector), the matrix-vector multiplication cannot be interpreted as a weighted sum in the same way. In this case, the dot product of a row of the matrix with the row vector would yield a scalar, not a vector.

Therefore, it is important to use the correct matrix-vector multiplication format depending on how the coordinates of the points are represented.
(2 votes)

Video transcript

- [Instructor] In this video, we're going to explore how a two by two matrix can be interpreted as representing a transformation on the coordinate plane. So let's just start with some examples or some conceptual ideas. So the first conceptual idea is that any point on our coordinate plane here and this of course is our X axis and this is our Y axis, can be represented by a combination of two vectors. You could have this vector right over here. That goes exactly one unit in the horizontal direction. We can represent that as a vector like this, one, zero, when you write a vector vertically like this the convention is that the top number here is what we're doing in the X direction. And then the bottom number, the zero is what we're doing in the vertical direction or the Y direction, so this is the one, zero vector. And then this right over here, what would we call that? Well that we would call the zero, one vector because it doesn't go at all in the X direction. And it only goes up one in the Y direction. Now just to feel good that any point on the coordinate plane can be represented as a weighted sum of these. Let's just pick a point at random. Let's say this point, all right over here, let's call it point A. And you could represent that as a vector that looks something like this, I'll do it as a dotted line, but this could be represented as a positional vector like that. And of course, if we're thinking about it in coordinates we would just say this is at the coordinate three comma one. The X coordinate is a three, the Y coordinate is one. But if we wanted to express it in terms of a vector, we could write it out as three, one, the X direction we're moving three from the origin, positive three and the Y direction we're moving one to get there. And you could see that we can represent this as a weighted sum of these two vectors. We can write this as, this is the same thing as three times our one, zero vector, one, zero plus one times our zero one, vector zero one vector. And you can see it visually, this yellow vector that points to point a right over there. You can have three of this vector, one, two, and then three. And then one of the orange vector. Now I said that I would explain how two by two matrices can represent a transformation. And the way that you could think about it is if I have a two by two matrix that looks like this, so let me just draw the matrix, where the first column is one, zero and then the second column is zero, one. This just tells you what to do with these two vectors. I know this might be a little bit confusing at first, but let's just walk through it together. So the way I've represented it, this first column says what is the transformation you want to apply to this one, zero vector, this first blue vector? Well, we're just keeping it one zero. So we're not changing it, it's the same. One, zero vector here one, zero vector here. Likewise, zero, one vector here, zero, one vector here. So this two by two matrix actually represents what's sometimes known as the identity transformation. It maps any point on the coordinate plane back to itself. It doesn't change the points, but I'm showing this because now I'm going to show a two by two matrix that represents a non identity transformation. For example, let me draw the matrix again. So let's say I have the matrix, instead of one, zero here I'm going to write a two, one here instead of a zero, one here let me write a one, two. So in this transformation, what we're doing is we're turning this one, zero vector into a two, one vector. You're going to see what I'm talking about in a second. So what does a two, one vector look like? Let me do it in that color. Well, we go two in the X direction one in the Y direction. So it's going to look like this, it's going to look like that. And then what does a one, two vector look like? Well, it goes one in the X direction and then two in the Y direction, so it looks like this. And the way that this represents a transformation is that anything that was a weighted sum of the one, zero and the zero, one vectors originally, you can now view as a weighted sum of the two, one and the one, two vectors. And so we can now think of another point A prime, that's not going to be three of the ones zeros and one of the zero ones. We can think of it as, let me write it over here, as three of the two, ones plus one of the one twos, one of these one, two. So where would that put that now? What we're going to go, go three of the two, ones. So this is one of them, this is two of them and then this is three of them. And then I'm going to have one of the one, two vectors, this orange vector right over here. And so I'm going to have one of those. And so that will take me to A prime. This is my new point after the transformation, I go to A prime. So I've gone from this point A to this A prime. And this two by two matrix is telling us how to transform. Now we can do that with multiple points. Now, if I talk about a point here at the origin that was originally at the origin point B, well that's zero of the orange vector, zero of the blue vectors. So even after the transformation, it's going to be zero of the two, one vector and zero of the one, two vectors. So it's just going to stay in place. It's just going to map to itself. So B is equal to B prime, and we could also imagine another point. Let's say right over here, let's call that point C. Well point C is originally two of the blue vectors and none of the orange vectors. So after the mapping, it'll be two of the two, one vectors and none of the one, two vectors. So two of the two ones. So if you go one, two, and then none of the one, twos you're going to get C prime right over there. And so notice, if originally you had a triangle between A, B and C, let me draw it like this. So originally you had this triangle A, B, C what is it now gotten mapped, what is it now mapped to? Well, it's now mapped to big triangle. I'll do my best to draw it relatively straight so we land on target. So that's that side. And then we have this side going from B prime to C prime and then we wanna connect that side from C prime to A prime. Now you might be saying, so how do you know that the lines map onto other lines? How do you know this transformation didn't all of a sudden make this line squiggly or zigzag? And that's one of the interesting properties of the type of transformation we're talking about. A two by two matrix will represent a linear transformation. And there's two ways to think about it in this context, a linear transformation will always map the origin onto itself and it will always map a line onto another line, it won't turn that line into a curve, or it won't make it zigzag somehow. Now the last thing you might be wondering is, Hey what about all these transformations we had from geometry, these similarity transformations, things like rotations reflections, dilations, can you do those with matrices? And the simple answer is yes, you can do them as long as you keep the origin in place. And you can actually using a two by two matrix come up with a whole series of other linear transformations that are much more, let's call it exotic, than just the rotations, reflections and dilations.