Why does reconstruction formula have the same formula as the projection of a vector onto a subspace? $$v = \sum_{i=1}^{n} {\langle v | e_i\rangle\, e_i}$$ where ${e_1, ..., e_n}$ is an orthonormal basis and v is in an inner product space$(V,\langle\,\cdot\,|\,\cdot\,\rangle)$. $V$ can be any vector space such as $\mathbb R^n$

  • 2
    $\begingroup$ Please state the formulas you are referring to. Do you mean orthogonal projection in a unitary/euclidean space when an orthonormal basis is given? $\endgroup$ – Christoph May 18 '18 at 7:20
  • $\begingroup$ I have now included the formula. Thanks. $\endgroup$ – Hypervoi3 May 18 '18 at 8:53

Let $U\subseteq V$ be a subspace of a vector space $U$. To define a projection onto $U$, one needs to choose a complement $W$. Here a complement is a second subspace $W\subseteq V$ such that $V=U\oplus W$, which means $V=U+W$ and $U\cap W=\{0\}$. In this situation, every $v\in V$ has a unique decomposition as $v=u+w$ with $u\in U$ and $w\in W$. Hence, we can define the projection onto $U$ along $W$ by $p(u+w)=u$.

Now we consider the case of a finite dimensional euclidean space and orthogonal projections. For a subspace $U\subseteq V$, the orthogonal projection is the projection onto $U$ along $U^\perp$, the orthogonal complement of $U$. This works since $V=U\oplus U^\perp$. Now to obtain the formula you are referring to, let $(u_1,\dots,u_k)$ be an orthonormal basis of $U$ and $(w_1,\dots,w_l)$ be an orthonormal basis of $U^T$. Together these form an orthonormal basis $(u_1,\dots,u_k,w_1,\dots,w_l)$ of the whole space $V$.

Given any vector $v\in V$ we can write it uniquely as a linear combination $$v = \sum_{i=1}^k \lambda_i u_i + \sum_{i=1}^l \mu_i w_i.$$ Furthermore, since we have an orthonormal basis, the coefficients are given by $\lambda_i = \langle v, u_i\rangle$ and $\mu_i = \langle v, w_i\rangle$. So we have the unique decomposition $$ v = \underbrace{\sum_{i=1}^k \langle v, u_i\rangle\, u_i}_{\in U} + \underbrace{\sum_{i=1}^l \langle v,w_i\rangle w_i}_{\in U^\perp}. $$ But this means that the projection of $V$ onto $U$ along $U^\perp$ is given by $$ p(v) = p\left(\underbrace{\sum_{i=1}^k \langle v, u_i\rangle\, u_i}_{\in U} + \underbrace{\sum_{i=1}^l \langle v,w_i\rangle w_i}_{\in U^\perp}\right) = \sum_{i=1}^k \langle v, u_i\rangle\, u_i. $$

| cite | improve this answer | |

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.