Download Change of Basis: Coordinate Vectors and Transition Matrices and more Exams Mathematics in PDF only on Docsity!
Change of Basis
Coordinate vectors. This are notes covering changing bases/coordinates. Here is the setup for all of the problems. We begin with a vector space V that has an ordered basis E = [v 1 ,... , vn]. (We always keep the same order for vectors in the basis.) By the “coordinate theorem,” if v ∈ V , then we can always express v ∈ V in one and only one way as a linear combination of the the vectors in E. Specifically, for any v ∈ V there are scalars x 1 ,... , xn such that v = x 1 v 1 + x 2 v 2 + · · · + xnvn. (1)
Moreover, by the same theorem, we have an isomorphism between V and Rn:
[v]E =
x 1 .. . xn
The column vector [v]E is the coordinate vector of v relative to E, and the xj ’s are the coordinates of v. The isomorphism just means that these properties hold:
[v + w]E = [v]E + [w]E and [cv]E = c[v]E. (2)
Examples. Let V = P 3 and E = { 1 , x, x^2 }. What is the coordinate vector [5 + 3x − x^2 ]B? Answer:
[5 + 3x − x^2 ]E =
If we ask the same question for [5 − x^2 + 3x]E , the answer is the same, because to find the coordinate vector we have to order the basis elements so that they are in the same order as E. Let’s turn the question around. Suppose that we are given
[p]E =
then what is p? Answer: p(x) = 3 · 1 + 0 · x + (−4) · x^2 = 3 − 4 x^2. Let’s try another space. Let V = span{ex, e−x}, which is a subspace of C(−∞, ∞). Here, we will take E = {ex, e−x}. What are coordinate
vectors for sinh(x) and cosh(x)? Solution: Since sinh(x) = 12 et^ − 12 e−x^ and cosh(x) = 12 ex^ + 12 e−x, these vectors are
[sinh(x)]E =
2 − (^12)
and [cosh(x)]E =
2 1 2
Change of basis Suppose that F = [w 1 ,... , wn] is a second basis for V. How do we relate coordinates relative to E to those of F? How do we change from one set of coordinates to another? Suppose that we know the E-coordinate vector of v, x := [v]E and we want F -coordinate vector y := [v]F. To do this, we first find the F - coordinates of the E-basis vectors; these are just the column vectors below.
s 1 = [v 1 ]F , s 2 = [v 2 ]F ,... , sn = [vn]F.
Now, in the representation (1) for v in the E-basis, take the F -coordinates of both sides to get
y = [v]F = [x 1 v 1 + x 2 v 2 + · · · + xnvn]F.
Using the properties of coordinate vectors in (2), we see that
y = [v]F = x 1 [v 1 ]F + x 2 [v 2 ]F + · · · + xn[vn]F = x 1 s 1 + x 2 s 2 + · · · + xnsn = [s 1 s 2 · · · sn] ︸ ︷︷ ︸ S
x = Sx
The matrix S is the transition matrix from E-coordinates to F -coordinates. When we want to emphasize this, we will write SE→F , instead of just S. Examples. Start out with V = P 3. Let E = [x + 1, x − 1 , 1 + x + x^2 ] and let F = [1, x, x^2 ]. To find the change of basis matrix SE→F , we need the F coordinate vectors for the E basis. These are easy to find.
s 1 = [1+x]F =
, s 2 = [x−1]F =
, s 3 = [1+x+x^2 ]F =
By what we said above, the transition matrix is
SE→F =
Change of basis for matrices for linear transformations. The ma- trix A that represents a linear transformation L : V → V relative to a basis E = [v 1 ,... , vn] is
A =
[
[L(v 1 )]E [L(v 2 )]E · · · [L(vn)]E
]
The equation L(v) = w is completely equivalent to the matrix equation A[v]E = [w]E. Of course, we could have used another basis F = [w 1 , w 2 ,... , wn], with another matrix
B =
[
[L(w 1 )]F [L(w 2 )]F · · · [L(wn)]F
]
representing L. If we know A and want to find B, start with A[v]E = [w]E. Use [v]E = SF toE [v]F to get ASF →E [v]F = [w]E. Then, note that [w]F = SE→F [w]E. Combining this with the previous equation gives SF →E ASF →E [v]F = [w]F. Letting S = SF →E , we obtain this relation between A and B: B = S−^1 AS An example. Let V = P 3 and take E = [1, x, x^2 ] and F = [1, x, 12 (3x^2 − 1)]. The linear transformation for this example is L(p) = ((x^2 − 1)p′)′. To find the matrix A that represents L, we first apply L to each of the basis vectors in B.
L(1) = 0, L(x) = 2x, and L(x^2 ) = −2 + 6x^2.
Next, we find the E-basis coordinate vectors for each of these.
[0]E =
(^) [2x]E =
(^) [−2 + 6x^2 ]E =
and so them matrix that represents L is
A =
We have already calculated the transition matrices, which are given in (4) and (5). Thus, since B = SE→F ASS→E , we have
B =
