2

If I take a vector v = [a,b], then isn't v.Mv = a^2, which is strictly greater than zero for all a and b not equal to zero?

Jeff
  • 21

3 Answers3

2

Positive definite means $v^tMv > 0$ for all $v \neq 0$. Not so for your matrix $M$. E.g., $v^t = (0, 1)$.

Simon S
  • 26,524
1

What you say is true. Yet the condition is for every non-zero vector. Note that $a$ can be $0$ while $v$ is not the zero-vector (when $b\neq 0$).

quid
  • 42,135
0

For a matrix to be positive definite, you need $\mathbf{v}^TM\mathbf{v}>0$ for any $\mathbf{v}\neq\mathbf{0}$.

An arbitrary vector $\mathbf{v}=(a,b)$ will indeed give you $\mathbf{v}^TM\mathbf{v}=a^2$, but $a^2\ge 0$, i.e it is not necessary that $a^2\neq 0$. You could for instance pick the vector $\mathbf{v}^T=(0,1)$.

However, you matrix is positive semi-definite, since you have $\mathbf{v}^TM\mathbf{v}\ge 0$.

Demosthene
  • 5,420