# Linear transformations and matrices | Essence of linear algebra, chapter 3 Unfortunately, no one can be told, what the
Matrix is. You have to see it for yourself.
– Morpheus Surprisingly apt words on the importance of
understanding matrix operations visually Hey everyone! If I had to choose just one topic that makes all of the others in linear algebra start to click and which too often goes unlearned the first
time a student takes linear algebra, it would be this one:
the idea of a linear transformation and its relation to matrices. For this video, I’m just going to focus on
what these transformations look like in the case of two dimensions and how they relate to the idea of matrix-vector
multiplication. In particular, I want to show you a way to
think about matrix-vector multiplication that doesn’t rely on memorization. To start, let’s just parse this term “linear
transformation”. “Transformation” is essentially a fancy
word for “function”. It’s something that takes in inputs and spits
out an output for each one. Specifically in the context of linear algebra,
we like to think about transformations that take in some vector and spit out another vector. So why use the word “transformation” instead
of “function” if they mean the same thing? Well, it’s to be suggestive of a certain way to
visualize this input-output relation. You see, a great way to understand functions
of vectors is to use movement. If a transformation takes some input vector
to some output vector, we imagine that input vector moving over to
the output vector. Then to understand the transformation as a
whole, we might imagine watching every possible input
vector move over to its corresponding output vector. It gets really crowded to think about all
of the vectors all at once, each one is an arrow, So, as I mentioned last video, a nice trick
is to conceptualize each vector, not as an arrow, but as a single point: the point where its
tip sits. That way to think about a transformation taking
every possible input vector to some output vector, we watch every point in space moving to some
other point. In the case of transformations in two dimensions, to get a better feel for the whole “shape”
of the transformation, I like to do this with all of the points on
an infinite grid. I also sometimes like to keep a copy of the
grid in the background, just to help keep track of where everything
ends up relative to where it starts. The effect for various transformations, moving
around all of the points in space, is, you’ve got to admit, beautiful. It gives the feeling of squishing and morphing
space itself. As you can imagine, though arbitrary transformations
can look pretty complicated, but luckily linear algebra limits itself to
a special type of transformation, ones that are easier to understand, called
“linear” transformations. Visually speaking, a transformation is linear
if it has two properties: all lines must remain lines, without getting
curved, and the origin must remain fixed in place. For example, this right here would not be
a linear transformation since the lines get all curvy and this one right here, although it keeps
the line straight, is not a linear transformation because it
moves the origin. This one here fixes the origin and it might
look like it keeps line straight, but that’s just because I’m only showing the
horizontal and vertical grid lines, when you see what it does to a diagonal line,
it becomes clear that it’s not at all linear since it turns that line all curvy. In general, you should think of linear transformations
as keeping grid lines parallel and evenly spaced. Some linear transformations are simple to
think about, like rotations about the origin. Others are a little trickier to describe with
words. So how do you think you could describe these
transformations numerically? If you were, say, programming some animations
to make a video teaching the topic what formula do you give the computer so that
if you give it the coordinates of a vector, it can give you the coordinates of where that
vector lands? It turns out that you only need to record where the two basis vectors, i-hat and j-hat, each land. and everything else will follow from that. For example, consider the vector v with coordinates
(-1,2), meaning that it equals -1 times i-hat + 2
times j-hat. If we play some transformation and follow
where all three of these vectors go the property that grid lines remain parallel
and evenly spaced has a really important consequence: the place where v lands will be -1 times the
vector where i-hat landed plus 2 times the vector where j-hat landed. In other words, it started off as a certain
linear combination of i-hat and j-hat and it ends up is that same linear combination
of where those two vectors landed. This means you can deduce where v must go
based only on where i-hat and j-hat each land. This is why I like keeping a copy of the original
grid in the background; for the transformation shown here we can read
off that i-hat lands on the coordinates (1,-2). and j-hat lands on the x-axis over at the
coordinates (3, 0). This means that the vector represented by
(-1) i-hat + 2 times j-hat ends up at (-1) times the vector (1, -2) +
2 times the vector (3, 0). Adding that all together, you can deduce that
it has to land on the vector (5, 2). This is a good point to pause and ponder,
because it’s pretty important. Now, given that I’m actually showing you the full transformation, you could have just looked to see the v has
the coordinates (5, 2), but the cool part here is that this gives
us a technique to deduce where any vectors land, so long as we have a record of where i-hat
and j-hat each land, without needing to watch the transformation
itself. Write the vector with more general coordinates
x and y, and it will land on x times the vector where
i-hat lands (1, -2), plus y times the vector where j-hat lands
(3, 0). Carrying out that sum, you see that it lands
at (1x+3y, -2x+0y). I give you any vector, and you can tell me
where that vector lands using this formula what all of this is saying is that a two dimensional
linear transformation is completely described by just four numbers: the two coordinates for where i-hat lands and the two coordinates for where j-hat lands. Isn’t that cool? it’s common to package these coordinates into a two-by-two grid of numbers, called a two-by-two matrix, where you can interpret the columns as the two special vectors where i-hat and j-hat each land. If you’re given a two-by-two matrix describing
a linear transformation and some specific vector and you want to know where that linear transformation
takes that vector, you can take the coordinates of the vector multiply them by the corresponding columns
of the matrix, then add together what you get. This corresponds with the idea of adding the
scaled versions of our new basis vectors. Let’s see what this looks like in the most
general case where your matrix has entries a, b, c, d and remember, this matrix is just a way of
packaging the information needed to describe a linear transformation. Always remember to interpret that first column,
(a, c), as the place where the first basis vector
lands and that second column, (b, d), is the place
where the second basis vector lands. When we apply this transformation to some
vector (x, y), what do you get? Well, it’ll be x times (a, c) plus y times (b, d). Putting this together, you get a vector (ax+by,
cx+dy). You can even define this as matrix-vector
multiplication when you put the matrix on the left of the
vector like it’s a function. Then, you could make high schoolers memorize
this, without showing them the crucial part that
makes it feel intuitive. But, isn’t it more fun to think about these columns as the transformed versions of your basis
vectors and to think about the results as the appropriate linear combination of those
vectors? Let’s practice describing a few linear transformations
with matrices. For example, if we rotate all of space 90° counterclockwise then i-hat lands on the coordinates (0, 1) and j-hat lands on the coordinates (-1, 0). So the matrix we end up with has columns
(0, 1), (-1, 0). To figure out what happens to any vector after
90° rotation, you could just multiply its coordinates by
this matrix. Here’s a fun transformation with a special
name, called a “shear”. In it, i-hat remains fixed so the first column of the matrix is (1, 0), but j-hat moves over to the coordinates (1,1) which become the second column of the matrix. And, at the risk of being redundant here, figuring out how a shear transforms a given
vector comes down to multiplying this matrix by that
vector. Let’s say we want to go the other way around, starting with the matrix, say with columns
(1, 2) and (3, 1), and we want to deduce what its transformation
looks like. Pause and take a moment to see if you can
imagine it. One way to do this is to first move i-hat to (1, 2). Then, move j-hat to (3, 1). Always moving the rest of space in such a
way that keeps grid lines parallel and evenly
spaced. If the vectors that i-hat and j-hat land on
are linearly dependent which, if you recall from last video, means that one is a scaled version of the
other. It means that the linear transformation squishes
all of 2D space on to the line where those two vectors sit, also known as the one-dimensional span of those two linearly dependent vectors. To sum up, linear transformations are a way to move around space such that the grid lines remain parallel and
evenly spaced and such that the origin remains fixed. Delightfully, these transformations can be described using
only a handful of numbers. The coordinates of where each basis vector
lands. Matrices give us a language to describe these
transformations where the columns represent those coordinates and matrix-vector multiplication is just a
way to compute what that transformation does to a given vector. The important take-away here is that, every time you see a matrix, you can interpret it as a certain transformation
of space. Once you really digest this idea, you’re in a great position to understand linear
algebra deeply. Almost all of the topics coming up, from matrix multiplication to determinant, change of basis, eigenvalues, … all of these will become easier to understand once you start thinking about matrices as
transformations of space. Most immediately, in the next video I’ll be talking about multiplying two matrices together. See you then!

1. Epic!!

2. This made me realize why the identity matrix is composed of 1s in the main diagonal, as those 1s mean the basic vectors are being scaled by 1 and only in its original direction, thus staying the same. Great video !

3. How do we know what the original matrix is that is being transformed? Do we compute this from the vector that we are transforming?

4. Oh my god, I instinctively hit the pause at 5.12 only to find out that he really tells us to pause at 5.14….

5. These animations are brilliant!!

6. Thank you sir…listening with more curiousity….thank you again!!!