video_id
stringclasses
7 values
text
stringlengths
2
29.3k
ZK3O402wf1c
How could it go wrong that out of these --
ZK3O402wf1c
out of three columns and all their combinations --
ZK3O402wf1c
when would I not be able to produce some b off here?
ZK3O402wf1c
When could it go wrong?
ZK3O402wf1c
Do you see that the combinations --
ZK3O402wf1c
let me say when it goes wrong.
ZK3O402wf1c
If these three columns all lie in the same plane,
ZK3O402wf1c
then their combinations will lie in that same plane.
ZK3O402wf1c
So then we're in trouble.
ZK3O402wf1c
If the three columns of my matrix --
ZK3O402wf1c
if those three vectors happen to lie in the same plane --
ZK3O402wf1c
for example, if column three is just
ZK3O402wf1c
the sum of column one and column two, I would be in trouble.
ZK3O402wf1c
That would be a matrix A where the answer would be no,
ZK3O402wf1c
because the combinations --
ZK3O402wf1c
if column three is in the same plane as column one and two,
ZK3O402wf1c
I don't get anything new from that.
ZK3O402wf1c
All the combinations are in the plane and only right-hand sides
ZK3O402wf1c
b that I could get would be the ones in that plane.
ZK3O402wf1c
So I could solve it for some right-hand sides, when
ZK3O402wf1c
b is in the plane, but most right-hand sides
ZK3O402wf1c
would be out of the plane and unreachable.
ZK3O402wf1c
So that would be a singular case.
ZK3O402wf1c
The matrix would be not invertible.
ZK3O402wf1c
There would not be a solution for every b.
ZK3O402wf1c
The answer would become no for that.
ZK3O402wf1c
Okay.
ZK3O402wf1c
I don't know --
ZK3O402wf1c
shall we take just a little shot at thinking
ZK3O402wf1c
about nine dimensions?
ZK3O402wf1c
Imagine that we have vectors with nine components.
ZK3O402wf1c
Well, it's going to be hard to visualize those.
ZK3O402wf1c
I don't pretend to do it.
ZK3O402wf1c
But somehow, pretend you do.
ZK3O402wf1c
Pretend we have -- if this was nine equations and nine
ZK3O402wf1c
unknowns, then we would have nine columns,
ZK3O402wf1c
and each one would be a vector in nine-dimensional space
ZK3O402wf1c
and we would be looking at their linear combinations.
ZK3O402wf1c
So we would be having the linear combinations
ZK3O402wf1c
of nine vectors in nine-dimensional space,
ZK3O402wf1c
and we would be trying to find the combination that hit
ZK3O402wf1c
the correct right-hand side b.
ZK3O402wf1c
And we might also ask the question can we always do it?
ZK3O402wf1c
Can we get every right-hand side b?
ZK3O402wf1c
And certainly it will depend on those nine columns.
ZK3O402wf1c
Sometimes the answer will be yes --
ZK3O402wf1c
if I picked a random matrix, it would be yes, actually.
ZK3O402wf1c
If I used MatLab and just used the random command, picked
ZK3O402wf1c
out a nine by nine matrix, I guarantee it would be
ZK3O402wf1c
good.
ZK3O402wf1c
It would be non-singular, it would
ZK3O402wf1c
be invertible, all beautiful.
ZK3O402wf1c
But if I choose those columns so that they're not independent,
ZK3O402wf1c
so that the ninth column is the same as the eighth column,
ZK3O402wf1c
then it contributes nothing new and there
ZK3O402wf1c
would be right-hand sides b that I couldn't get.
ZK3O402wf1c
Can you sort of think about nine vectors
ZK3O402wf1c
in nine-dimensional space an take their combinations?
ZK3O402wf1c
That's really the central thought --
ZK3O402wf1c
that you get kind of used to in linear algebra.
ZK3O402wf1c
Even though you can't really visualize it,
ZK3O402wf1c
you sort of think you can after a while.
ZK3O402wf1c
Those nine columns and all their combinations
ZK3O402wf1c
may very well fill out the whole nine-dimensional space.
ZK3O402wf1c
But if the ninth column happened to be the same as the eighth
ZK3O402wf1c
column and gave nothing new, then probably what it would
ZK3O402wf1c
fill out would be --
ZK3O402wf1c
I hesitate even to say this -- it would be a sort of a plane
ZK3O402wf1c
--
ZK3O402wf1c
an eight dimensional plane inside nine-dimensional space.
ZK3O402wf1c
And it's those eight dimensional planes
ZK3O402wf1c
inside nine-dimensional space that we
ZK3O402wf1c
have to work with eventually.
ZK3O402wf1c
For now, let's stay with a nice case where the matrices work,
ZK3O402wf1c
we can get every right-hand side b and here
ZK3O402wf1c
we see how to do it with columns.
ZK3O402wf1c
Okay.
ZK3O402wf1c
There was one step which I realized
ZK3O402wf1c
I was saying in words that I now want to write in letters.
ZK3O402wf1c
Because I'm coming back to the matrix form of the equation,
ZK3O402wf1c
so let me write it here.
ZK3O402wf1c
The matrix form of my equation, of my system
ZK3O402wf1c
is some matrix A times some vector x
ZK3O402wf1c
equals some right-hand side b.
ZK3O402wf1c
Okay.
ZK3O402wf1c
So this is a multiplication.
ZK3O402wf1c
A times x.
ZK3O402wf1c
Matrix times vector, and I just want to say
ZK3O402wf1c
how do you multiply a matrix by a vector?
ZK3O402wf1c
Okay, so I'm just going to create a matrix --
ZK3O402wf1c
let me take two five one three --
ZK3O402wf1c
and let me take a vector x to be, say, 1and 2.
ZK3O402wf1c
How do I multiply a matrix by a vector?
ZK3O402wf1c
But just think a little bit about matrix notation
ZK3O402wf1c
and how to do that in multiplication.
ZK3O402wf1c
So let me say how I multiply a matrix by a vector.
ZK3O402wf1c
Actually, there are two ways to do it.
ZK3O402wf1c
Let me tell you my favorite way.
ZK3O402wf1c
It's columns again.
ZK3O402wf1c
It's a column at a time.