0
$\begingroup$

Question:

Suppose that $V$ is an $m$-dimensional affine subspace of $\mathbb R^n$, with $m < n$. show that there exist linearly independent vectors $a_1, \dots, a_{n-m}$, and scalars $b_1, \dots , b_{n-m}$ such that $$ V = \{y : a_i'y = b_i, \quad i = 1,...,n-m\} $$

Where I am: So I know that we are working with a subspace that if represented in matrix form has less rows than columns, i.e. one or more dimension of the subspace could be unbounded. I also know that an affine subspace would be like a subspace that doesn't include the origin (it has been shifted away from another parallel subspace). So using all of this, to my best knowledge: I need to show that there is an Ax = b equation where all the vectors a$_i$ that make up A are linearly independent but not zero!! (Affine subspace)

Where to go with this though?? I was thinking about playing around with the vectors (2,1,1) and (1,2,1) which (I think) satisfy the the linear independent and affine requirements, but I am not sure.

Any help here would be much appreciated. I am not a wise man, and sorry for the horibly phrased question! :C

$\endgroup$
1
$\begingroup$

Your intuition is good: if $V \subset \mathbb R^n$ is an $m$-dimensional affine space, then there exists some vector $t \in \mathbb R^n$ such that the translation $V - t =: U$ is an $m$-dimensional linear subspace of $\mathbb R^n$. We may pick an orthonormal basis $u_1, \dots, u_m$, of $U$, and this extends to an orthonormal basis $u_1, \dots, u_m, a_1, \dots, a_{n-m}$ of $V$. As described, we have $u \perp a_i$ for every $u \in U$, i.e., $$ U = \{ u \in \mathbb R^n : a_i' u = 0, \quad i = 1, \dots, n-m\}, $$ where $a_i'$ denotes the transpose of the column vector $a_i$. Then \begin{align*} V &= U + t \\ &= \{u + t \in \mathbb R^n : a_i' u = 0 \quad\forall i \in \{1, \dots, n-m\}\} \\ &= \{v \in \mathbb R^n : a_i' (v - t) = 0 \quad\forall i \in \{1, \dots, n-m\}\} \\ &= \{v \in \mathbb R^n : a_i' v = a_i' t \quad\forall i \in \{1, \dots, n-m\} \} && \text{substitute } v = u+t \end{align*} and the desired scalars are $b_i = a_i ' t$.

$\endgroup$
  • $\begingroup$ awesome! Thank you so much, this is very clear to me, way more so than my textbook. $\endgroup$ – wbrugato Apr 14 '16 at 14:56

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.