1
$\begingroup$

Given $d+2$ vectors $x_1,x_2,\dots,x_{d+2} \in \mathbb{R}^d$. Clearly, they are linearly dependent. However, we have better result: Let $$ y_i = \begin{bmatrix} x_i\\ 1 \end{bmatrix} \in \mathbb{R}^{d+1}. $$ They are also linearly dependent. Thus, there exists coefficients $a_1, a_2, \dots, a_{d+2}$ that are not simultaneously zero which satisfy $a_1y_1 + a_2y_2 + \dots + a_{d+2}y_{d+2} = 0$ (or in other words, $a_1x_1+a_2x_2+\dots+a_{d+2}x_{d+2}=0$ and $a_1+a_2+\dots+a_{d+2}=0$).

Question is to determine sign of $a_i$. Obviously, it can be anything since we can just multiply all of the coefficients by $-1$. Nevertheless, there is some relation between sign of $a_i$ and sign of determinant of $$ \begin{bmatrix} y_1 & \cdots & y_{i-1} & y_{i+1} & \cdots & y_{d+2} \end{bmatrix} \in \mathbb{R}^{(d+1) \times (d+1)}. $$ For every $1 \leq i \leq d+2$, $$ \frac{\operatorname{sign}(a_i)}{\operatorname{sign} \Bigl( \det \bigl( \begin{bmatrix} y_1 & \cdots & y_{i-1} & y_{i+1} & \cdots & y_{d+2} \end{bmatrix} \bigr) \Bigr)} $$ is same. Any idea on how to prove this result? Thanks!

$\endgroup$

1 Answer 1

2
$\begingroup$

For $i=1,\dots, d+1$, the $a_i's$ are the solutions of $$ (y_1, y_2,\dots, y_{d+1}) \left ( \begin{array}{c} a_1\\ a_2\\ \vdots\\ a_{d+1} \end{array} \right)=-a_{d+2}y_{d+2}$$ Using Cramer's rule for solving linear systems, the solution is given by $$\begin{split} a_i&=\frac{\det(y_1, \dots, y_{i-1}, -a_{d+2}y_{d+2},y_{i+1},\dots,y_{d+1})}{\det(y_1, \dots, y_{d+1})}\\ &= (-1)^{d-i}a_{d+2}\frac{\det(y_1, \dots, y_{i-1},y_{i+1},\dots, y_{d+2})}{\det(y_1, \dots, y_{d+1})} \end{split} $$ which gives the result but with a factor of $(-1)^i$. I'm not sure where I messed up.

$\endgroup$

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.