Scribd
Upload
8 views

Definitions

Uploaded by

me homer
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
8 views

Definitions

Uploaded by

me homer
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

MATH 51 — Linear Algebra Through the World of Robots

Jonathan Lee

February 2, 2010

The aim of this note is to give hopefully more intuitive definitions of some of the linear algebra
concepts covered in class. (Disclaimer: this will be edited more later, with more examples added,
as I’ve just started typing this at 1:00am. Let me know if there are things you’d like added.)
Our setup: we’re in n-dimensional space, also known as Rn , and we’ve placed a remote-controlled
robot named Karel at the origin. The remote control has k buttons, each labelled with a vector
from Rn — for the sake of notation, assume the i-th button is labelled with a given vector vi 2 Rn ,
where 1  i  k. Karel has been programmed to move, changing its position by the vector vi , each
time button i is pressed. (Note that the buttons are magically engineered, in the sense that each
can also be pushed a negative or fractional number of times.)
Example 1. Suppose Karel is at the position (3, 4), and a button labelled with (9, 2) is pushed
twice. Then, Karel’s new position will be

(3, 4) + 2 · (9, 2) = (21, 8) .

Definition 1 (Span — technical). The span of the vectors v1 , v2 , . . . , vk is the set of all possible
values of
c1 v1 + c2 v2 + · · · + ck vk ,
where the ci are arbitrarily chosen real numbers.

Definition 2 (Span — robot). The span of the vectors v1 , v2 , . . . , vk is the set of all points Karel
can reach by having buttons on the remote control pushed, assuming Karel starts at the origin.

Example 2. Suppose Karel is in R3 , currently at the origin. If the remote control has two buttons,
v1 = (1, 0, 0) and v2 = (0, 1, 0), then the set of points Karel can reach is the xy-plane (which is the
span of v1 and v2 ).

Definition 3 (Linear independence — technical). The vectors v1 , v2 , . . . , vk are linearly inde-


pendent if whenever c1 , c2 , . . . , ck are real numbers, not all zero, then the vector

c1 v1 + c2 v2 + · · · + ck vk

is also not zero.

Definition 4 (Linear independence — robot). The vectors v1 , v2 , . . . , vk are linearly independent


if there is no way of pushing the buttons (using at least one, but not necessarily all) such that Karel
ends up where it started.

1
Example 3. Suppose Karel lives in R2 , and there are three remote-control buttons:

v1 = (2, 0) , v2 = (0, 3) and v3 = (4, 12) .

Then, these vectors are linearly dependent, since no matter where Karel starts, it’ll return to its
starting position if we push button 1 once, button 2 twice and button 3 negative two times.
The two definitions above have formulations involving a single matrix, instead of numerous
vectors. Keeping our same vectors v1 , v2 , . . . , vk 2 Rn , let A be the n ⇥ k matrix whose i-th column
is the vector vi .
Definition 5 (Matrix-vector product — technical). Let x = [c1 c2 · · · ck ]T be a vector in Rk .
Then the matrix-vector product Ax is the column vector

c1 v1 + c2 v2 + · · · + ck vk .

Definition 6 (Matrix-vector product — robot). Let x = [c1 c2 · · · ck ]T be a vector in Rk . Suppose


Karel starts o↵ at the origin. Then, the matrix-vector product Ax is where Karel ends up after
pushing the first button c1 times, the second button c2 times, and so on.
Example 4. Let Karel live in R2 , and suppose the remote control has the same buttons as in the
previous example, so that 
2 0 4
A= .
0 3 12
If we push the first button once, the second button twice and the third button thrice, then Karel
ends up in position (14, 42). This is the same as saying that
0 1
1 ✓ ◆
@ A 14
A 2 = .
42
3

Column space is the matrix version of span:


Definition 7 (Column space — technical). The column space of a matrix is the span of its
columns.
Definition 8 (Column space — robot). Let Karel’s remote control buttons correspond to the
column vectors of a matrix A. Then, the column space of A is all points Karel can reach (when
starting at the origin) by pushing its buttons.
We can reword one of the previous examples as follows:
Example 5. Let 2 3
1 0
A=4 0 1 5.
0 0
Then, the column space of A is the xy-plane. One possible basis of the column space is {(1, 0, 0), (0, 1, 0)},
but there are also weirder ones, like {(3, 5, 0), (4, 2, 0)}.
Definition 9 (Nullspace — technical). The nullspace of a matrix A is all vectors v such that
Av = 0.

2
Definition 10 (Nullspace — robot). Let Karel have remote control buttons corresponding to the
matrix A. Then, the nullspace of A is all possible ways of pushing Karel’s buttons so that it
returns to its starting position. (Notationally, the vector (3, 5, 9, . . .) means push the first button 3
times, the second button 5 times, the third button 9 times, and so on.)

As before, we can reword a previous example:


Example 6. Let 
2 0 4
A= .
0 3 12
No matter where Karel starts, it’ll return to its starting position if we push button 1 once, button
2 twice and button 3 negative two times. In other words, (1, 2, 2) is in the nullspace of A. Other
vectors in the nullspace of A are (0, 0, 0), ( 3, 6, 6) and (⇡, 2⇡, 2⇡), for example. The nullspace
can be described as the line in R3 parametrized (traced out) by the path f (t) = (t, 2t, 2t); one
possible basis is the single vector (1, 2, 2), but any point on the line (other than the origin) will
do.

Definition 11 (Basis — technical). Let V be a subspace of Rn . The vectors v1 , v2 , . . . , vk are a


basis for V if they span V and are linearly independent.

Definition 12 (Basis — robot). Let V be a subspace of Rn , and suppose Karel’s remote control
buttons are labelled v1 , v2 , . . . , vk . Then these vectors form a basis for V if Karel cannot leave V ,
and furthermore, for any point in V , there is exactly one way of pushing Karel’s buttons so it ends
up there. (Thus, there is a correspondence between points in V and ways of pushing the buttons.)

You might also like