1
$\begingroup$

I'm trying a reproduce the result of a paper. The paper contains a formula:

argmax formula with succeed equal

X is a matrix. And this matrix should be "succeed equal" to 0.

succeed equal

What does this mean for the formula and for argmax? What does this compare operator mean in general?

  • 0
    Where have you seen the strange naming "succeed equal" ? Where is the success behind that ? Why not "backward looking eye" ? – 2017-01-14
  • 0
    @JeanMarie I had this naming from wolframalpha: http://reference.wolfram.com/language/ref/SucceedsEqual.html – 2017-01-14

2 Answers 2

1

We usually write the curly comparison $\succeq$ for more general comparison operations. For vectors $x \in \mathbb{R}^n$ we would write $x \succeq 0$ to mean that $x_1 \ge 0, ..., x_n \ge 0$. For matrices, we write $X \succeq 0$ to mean that $X$ is positive semi-definite. That is, all of it's eigenvalues are greater than or equal to zero. (In convex analysis we generalize further, it is common to define a general comparison operator $\succeq_K$ with respect to some cone $K$.)

The operation $\underset{X \succeq 0}{\text{argmax }} f(X)$ returns the matrix $Z$ such that $Z \succeq 0$ and $f(Z) \ge f(Y)$ for any other $Y \succeq 0$. That is, we optimize $f$ over the space of positive semidefinite matrices.

3

It usually means that $X$ is positive semidefinite.