Let $\langle x,y\rangle$ be an inner product on $F^m$ and define $\|x\| = \sqrt{\langle x,x\rangle}$. Suppose $W$ is a subspace of $F^m$ and $\dim(W) = n$.

Let $β = \{v_1, v_2, \ldots , v_n\}$ be an orthogonal basis for $W$. Prove that for any vector $v ∈ W$ we have $v = \frac{\langle v,v_1\rangle}{||v_1||^2} v_1 + \cdots + \frac{\langle v,v_n\rangle}{\|v_n\|^2} v_n$.

I recognize this as the projection of $v_1$ onto $v +{}$ the projection of $v_2$ onto $v + \cdots +{}$ the projection of $v_n$ onto $v$ but I'm not sure how to show $v$ is equal to this. I'm assuming that because $β$ is a basis I can write any vector $v$ as a linear combination of its vectors (i.e. $v = c_1v_1 + \cdots + c_nv_n$), but then I don't know if I can show that $c_k = \frac{\langle v,v_k\rangle}{\|v_k\|^2}$ for every $k$. I think there's something about the basis being orthogonal that can help me reach my conclusion, but I can't seem to figure out exactly how to use this fact to help me.

  • $\begingroup$ I'm the \langle \rangle fairy, here to let you know that $\langle, \rangle$ plays nicer with TeX than <, > does :) $\endgroup$ – Patrick Stevens Mar 23 '17 at 8:00
  • $\begingroup$ Thanks! Will use that in the future! $\endgroup$ – Chris Mar 23 '17 at 17:08
  • $\begingroup$ Note proper MathJax usage, as in my edits to this question. $\langle a, b\rangle$ is right; $<a,b>$ is not. Also $\|a\|$ rather than $||a||,$ $a+\cdots+z$ rather than $a+ ... + z,$ and $\dim(W)$ rather than $dim(W).\qquad$ $\endgroup$ – Michael Hardy Feb 3 '18 at 17:51

Denoting $v = c_1v_1 + c_2v_2 + \cdots + c_nv_n$ for some $c_1, c_2, \cdots, c_n$ since $\{v_1, v_2, \cdots, v_n\}$ is a basis. We have \begin{align} & v - \frac{\langle v, v_1\rangle}{\|v_1\|^2}v_1 - \cdots - \frac{\langle v, v_n\rangle}{\|v_n\|^2}v_n \\ =\ &\left(c_1 - \frac{\langle v, v_1\rangle}{\|v_1\|^2}\right)v_1 + \cdots + \left(c_n - \frac{\langle v, v_n\rangle}{\|v_n\|^2}\right)v_n \\ =\ & \left(c_1 - \frac{\langle c_1v_1, v_1\rangle}{\|v_1\|^2}\right)v_1 + \cdots + \left(c_n - \frac{\langle c_nv_n, v_n\rangle}{\|v_n\|^2}\right)v_n \\ =\ &\left(c_1 - c_1\right)v_1 + \cdots + \left(c_n - c_n\right)v_n \\ =\ &0 \end{align} where the second equality is because $$ \langle v, v_i \rangle = \langle c_1v_1 + c_2v_2 + \cdots + c_nv_n, v_i\rangle = \langle c_iv_i, v_i\rangle = c_i \langle v_i, v_i \rangle = c_i\|v_i\|^2 $$

  • 2
    $\begingroup$ From $v = c_1v_1 + c_2v_2 + \cdots + c_nv_n$, taking inner product with $v_i$, we get $\langle v,v_i \rangle = c_i \langle v_i,v_i\rangle = c_i\|v_i\|^2$, since $\langle v_i , v_j \rangle = 0$ when $i \neq j$ $\endgroup$ – user348749 Mar 23 '17 at 7:49
  • $\begingroup$ @Muralidharan Thanks. It makes it more clear. $\endgroup$ – PSPACEhard Mar 23 '17 at 7:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.