Chapter 1 Introduction
1.1 Notations
- \(\inf\) and \(\sup\): similar to \(\min\) and \(\max\).
1.2 Generalized inequalities
The generalized inequality (defined by a proper cone \(K\)) is written as \[\begin{equation} \tag{1.1} x\preceq_{K} y \Longleftrightarrow y-x \in K \end{equation}\]
The componentwise inequality (\(K=R^n_+\)) is written as \[\begin{equation} \tag{1.2} x\preceq_{R_+^n} y \Longleftrightarrow x_i \leq y_i \end{equation}\]
The matrix inequality (\(K=S^n_+\)) is written as \[\begin{equation} \tag{1.3} X\preceq_{S_+^n} Y \Longleftrightarrow Y - X \text{ positive definite } \end{equation}\]
1.3 Minimum and minimal elements
- Minimum: all points are more
- Minimal: no points are less
1.4 Separating hyperplane theorem
Note that this is not strict (both inequalities includes equality).
1.4.1 Supporting hyperplane theorem
1.5 Dual cones and generalized inequalities
Examples:
- \(R^n_+\): \(R^n_+\)
- \(K = S_+^n\): \(S_+^n\) (\(<x,y> = Tr(XY)\))
- \(K = \{ (x, t) | \| x \|_2 \leq t \}\): \(K = \{ (x, t) | \| x \|_2 \leq t \}\)
- \(K = \{ (x, t) | \| x \|_1 \leq t \}\): \(K = \{ (x, t) | \| x \|_\infty \leq t \}\)
Note that \((K^*)^* = K\) only if \(K\) is proper.
Also, since dual cones of proper cones are proper, this defines generalized inequalities: \[\begin{equation*} y \succeq_K^* 0 \Longleftrightarrow y^Tx \geq 0 \text{ for all } x\succeq_K 0 \end{equation*}\]