Transformation matrix with respect to two bases?

Join the discussion
Ask a follow-up here, or get your own question answered by working scientists, mathematicians and engineers — people, not an autocomplete.
Real named experts · corrections over time · the nuance an AI answer skips
13 replies · 5K views
PhizKid
Messages
477
Reaction score
2

Homework Statement


Let ##S = \{1, e^x, e^{-x}, e^{2x}, e^{-2x}\}## and ##B = \{1, sinh(x),cosh(x), sinh(2x), cosh(2x)\}##. S spans the vector space V, and a linear transformation T: V -> V is defined by T(y) = y'' - 3y' - 4y.

(a) Find the representation matrix of T with respect to the bases S and B.
(b) Use the change of basis matrix (transition matrix) P, from S to B, to find the representation matrix of T with respect to the bases B and B.

The Attempt at a Solution


(a) I don't understand what it means to find a single transformation matrix for two different bases.

(b) I already found the change of basis matrix P from S to B by representing the elements in S in terms of the ones in B and set them as columns of the matrix P. But again, I don't understand what it means to use this change of basis to find a transformation matrix for two different bases, but this time the two bases are the same bases?
 
Physics news on Phys.org
You have two vectors related by ##\vec{y} = T(\vec{x})##. You're used to working with one basis to represent both ##\vec{x}## and ##\vec{y}## and find the matrix of T that relates the coordinate vectors, but you don't have to use the same basis for both. One vector could be represented relative to B while the other is relative to S. Since you'll have a different coordinate vector for one of the vectors, the matrix for T will be different.
 
Then I really don't understand the questions because it's asking for a transformation matrix with respect to both S and B, and then another transformation matrix with respect to B and also B again?
 
You want to find matrices TBB and TSB (or TBS — the question isn't clear). They're related by P.

TSB, for example, takes the coordinates of ##\vec{x}## wrt to basis B and gives you the coordinates of ##\vec{y}## wrt basis S.

TBB takes the coordinates of ##\vec{x}## wrt to basis B and gives you the coordinates of ##\vec{y}## wrt basis B.

Sorry...I have to run...I'll post a bit later if someone else already hasn't.
 
  • Like
Likes   Reactions: 1 person
Okay, for T(y) = y'' - 3y' - 4y, I found the matrix that takes S and transforms it into coordinates wrt S. I just took the derivatives of each element of S and plugged them into T(y) and found ##\{-4, -6e^x, 0, -6e^{2x}, 6e^{-2x}\}## so I just put each element in its respective positive in each column.

So for part (a), I want to take S, transform the basis into B by using P, and then somehow get a transformation for this new basis B, but I only have a transformation that takes me from S to S. What should I do here?
 
Let me use a simple example with two basis vectors. The two bases are S={sinh x, cosh x} and B={ex, e-x}. Suppose we have the linear transformation T(y) = y'.

cosh x = T(sinh x) would correspond to
\begin{align*}
\begin{pmatrix} 0 \\ 1 \end{pmatrix}_S &= T_{SS}\begin{pmatrix} 1 \\ 0 \end{pmatrix}_S \\
\begin{pmatrix} 1/2 \\ 1/2 \end{pmatrix}_B &= T_{BS}\begin{pmatrix} 1 \\ 0 \end{pmatrix}_S \\
\begin{pmatrix} 0 \\ 1 \end{pmatrix}_S &= T_{SB}\begin{pmatrix} 1/2 \\ -1/2 \end{pmatrix}_B \\
\begin{pmatrix} 1/2 \\ 1/2 \end{pmatrix}_B &= T_{BB}\begin{pmatrix} 1/2 \\ -1/2 \end{pmatrix}_B
\end{align*} depending on which basis you use for the domain and which basis you use for the range.
 
For the 2nd transformation you listed, how do I obtain that ##T_{BS}##? Is it the same thing as changing the basis S into basis B, then applying the transformation with respect to B to obtain the same answer?

I think I got part (b) by doing this: ##T_{BB} = PT_{SS}P^{-1}## where ##T_{SS}## is the transformation I listed above
 
You have
$$\sinh x = (1) \sinh x + (0) \cosh x = \left(\frac{1}{2}\right) e^x + \left(-\frac{1}{2}\right) e^{-x}$$ so with respect to the two bases, the representation for sinh x is ##\begin{pmatrix} 1 \\ 0 \end{pmatrix}## and ##\begin{pmatrix} 1/2 \\ -1/2 \end{pmatrix}##. You can do the same thing for cosh x. I just plugged in the various representations for sinh x and cosh x to get the four equations.

I'm not sure if this answers your question.
 
If you're asking how to get the matrix itself, the second equation tells you the first column of TBS is ##\begin{pmatrix} 1/2 \\ 1/2 \end{pmatrix}##. If you don't see this, just multiply in the vector ##\begin{pmatrix} 1 \\ 0 \end{pmatrix}## into a 2x2 matrix, and you'll see you get the first column. You can get the second column similarly, by multiplying by ##\begin{pmatrix} 0 \\ 1 \end{pmatrix}##.
 
For the second equation, you have ##(1,0)T_{BS} = (1/2, 1/2)##. I was asking how to obtain the general matrix ##T_{BS}## to directly transform something in S directly to some coordinates in B.
 
When something says a matrix is respect to 'B' how can you tell if it means with respect to 'S to B' or 'B to B'? It says that when T is a transformation with respect to 'S', ##PTP^{-1}## is = to the transformation with respect to B.
 
If only one basis is specified, you can assume both the domain and range are expressed in this basis.

In your example, matrix ##P## maps the representation of a vector wrt to S to its representation wrt B. ##P^{-1}## goes in the opposite direction. So you have
\begin{align*}
\vec{x}_S &= P^{-1}\vec{x}_B \\
\vec{y}_S &= T\vec{x}_S = TP^{-1}\vec{x}_B \\
\vec{y}_B &= P\vec{y}_S = PTP^{-1}\vec{x}_B
\end{align*} You can therefore identify ##PTP^{-1}## as the matrix of T with respect to B.
 
I am still having trouble understanding how to get the transformations. From your example here: https://www.physicsforums.com/showpost.php?p=4599971&postcount=6

The 3rd equation from the top. It means that some 2x2 transformation matrix multiplied on the left side of some arbitrary vector in B will equal to some arbitrary vector in S, correct? So my goal here is to find that 2x2 transformation matrix. However, in order to solve for this 2x2 transformation matrix, I need the inverse of a 2x1 matrix which is impossible. So how can I solve for the 2x2 matrix ##T_{SB}## in your example?