arXiv reader

Feasibility Conditions of Robust Portfolio Solutions under Single and Combined Uncertainties

Pulak Swain and Akshay Kumar Ojha

arXiv
2311.04609v1 · PDF · 8 November 2023
Categories
math.OC
Licence
CC BY 4.0 The site shows the text.

Abstract

In this paper, we derive the feasibility conditions for the robust counterparts of the uncertain Markowitz model. Our study is based on ellipsoidal, box, polyhedral uncertainty sets and also the uncertainty sets obtained from their combinations. We write the uncertain Markowitz problem in a quadratically constrained convex quadratic programming form and the uncertain sets are converted into their quadratic forms for the derivation. The feasibility conditions of robust solutions which we obtained are in the form of positive semi-definite matrices. The study can be useful for deriving the robust counterparts of the combined uncertainties, which in general is difficult to find out.

Items 12

Definition

Definition 1 (Robust Feasible Solution)

OK8P

[Ben-Tal, 2009] A vector \bm x is said to be the robust feasible solution of the uncertain problem (?) if it satisfies the uncertain constraints for all realizations of the uncertain set \mathscr{U}, that is, if \bm x satisfies

\begin{aligned} \begin{aligned} c(\bm{x},\bm{u}) \leq 0, \quad \forall \bm u \in \mathscr{U} \end{aligned} \end{aligned}
Used by Definition 3
Definition

Definition 2 (Robust Value)

19D9

[Ben-Tal, 2009] For a given candidate solution \bm{x}, the robust value \widehat{f}(\bm{x}) of the objective in problem (?) is the largest value of f(\bm{x},\bm{u}) over all realizations of the data from the uncertain set, that is

\begin{aligned} \begin{aligned} \widehat{f}(\bm{x})= \max_{\bm u \in \mathscr{U}} f(\bm{x},\bm{u}) \end{aligned} \end{aligned}
Used by Definition 3
Definition

Definition 3 (Robust Counterpart)

ABMZ

[Ben-Tal, 2009] The robust counterpart of the uncertain problem (?) is the optimization problem

\begin{aligned} \begin{aligned} & \min_{\bm x} && \left \{ \max_{\bm u \in \mathscr{U}} f(\bm x, \bm u) \right \}\\ &\text{s.t.:} && c(\bm x, \bm u) \leq \bm 0, \quad \forall \bm u \in \mathscr{U} \end{aligned} \end{aligned}

of minimizing the robust value of the objective over all robust feasible solutions to the uncertain problem.

Uses Definition 2, Definition 1 · Used by Definition 4, Theorem 0.1
Definition

Definition 4 (Robust Optimal Solution)

EOQ3

[Ben-Tal, 2009] The optimal solution of the robust counterpart problem (3) or is said to be the robust optimal solution of the uncertain problem (?).

Uses Definition 3
Theorem

Theorem 0.1

8N8T

[Ben-Tal, 1999] The robust counterpart of the uncertain problem (?) associated with the uncertain set \mathscr{U}=\mathscr{U}_1 \times \mathscr{U}_2 \times \dots \times \mathscr{U}_m, \mathscr{U}_i’s being given by (?), is equivalent to the SDP problem

\begin{aligned} \begin{aligned} &\min_x && \bm c ^\top \bm x \\ &\textrm{w.r.t.} && \bm x \in \mathbb{R}^n, \ \lambda^1, \dots, \lambda^n \in \mathbb{R}\\ & \textrm{s.t.:}&& \begin{small}\left [\begin{array} {c|c c c c|c} \gamma^{i(0)}+2 \bm x^\top\bm b^{i(0)}-\lambda^i & \dfrac{\gamma^{i(1)}}{2}+\bm x^\top \bm b^{i(1)} & \dfrac{\gamma^{i(1)}}{2}+\bm x^\top \bm b^{i(1)} & \dots&\dfrac{\gamma^{i(1)}}{2}+\bm x^\top \bm b^{i(1)} & [\bm A^{i(1)}\bm x]^\top\\ \hline \dfrac{\gamma^{i(1)}}{2}+\bm x^\top \bm b^{i(1)} & \lambda^i & & & & [\bm A^{i(1)}\bm x]^\top\\ \dfrac{\gamma^{i(2)}}{2}+\bm x^\top \bm b^{i(2)} & & \lambda^i & & & [\bm A^{i(2)}\bm x]^\top\\ \vdots & & &\ddots& & \vdots \\ \dfrac{\gamma^{i(k)}}{2}+\bm x^\top \bm b^{i(k)} & & & & \lambda^i & [\bm A^{i(k)}\bm x]^\top\\ \hline \bm A^{i(0)}\bm x & \bm A^{i(1)}\bm x & \bm A^{i(2)}\bm x & \dots & \bm A^{i(k)}\bm x & I_{l_i} \end{array}\right ]\end{small} \succeq 0\\ & && \text{for } i=1,2,\dots,m \text{ and $I_{l_i}$ is the identity matrix of order $l_i$}. \end{aligned} \end{aligned}
Uses Definition 3
Lemma

Lemma 1 (S-Lemma)

OJ4Z

[Ben-Tal, 2009] Let \bm A, \bm B be symmetric matrices of the same size such that, there exists \bm{\bar z} satisfying \bm{\bar z}^\top \bm A \bm{\bar z} \geq 0. Then the implication

\begin{aligned} \begin{aligned} \bm z^\top \bm A \bm z \geq 0 \implies \bm z^\top \bm B \bm z \geq 0 \end{aligned} \end{aligned}

holds true if and only if

\begin{aligned} \begin{aligned} \exists \ \lambda \geq 0 \quad : \quad \bm B \succeq \lambda \bm A \end{aligned} \end{aligned}
Theorem

Theorem 0.2

4GY6

Consider the uncertain Markowitz portfolio problem (?) under ellipsoidal uncertainty. The solution vector \bm x to the problem is feasible for the robust constraint in (?) if and only if there exists a \lambda \geq 0 such that the matrix

Scriptsize.
\begin{aligned} \begin{bmatrix} ({\bm{\mu}^{(0)}}^\top\bm x)^2-\tau^2-\lambda & ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x) & ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x) & \dots & ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(1)}}^\top\bm x)^2+\lambda & ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x) & \dots & ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x) & ({\bm{\mu}^{(2)}}^\top\bm x)^2+\lambda & \dots & ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ \vdots & \vdots & \vdots & \ddots & \vdots \\ ({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x) &({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x) & \dots & ({\bm{\mu}^{(n)}}^\top\bm x)^2+\lambda \end{bmatrix} \end{aligned}

is positive semidefinite.

Theorem

Theorem 0.3

MVVO

Consider the uncertain portfolio problem (?) under box uncertainty. The solution vector \bm x to the problem is feasible for the robust constraint in (?) if and only if there exists a \lambda \geq 0 such that the matrix

Scriptsize.
\begin{aligned} \begin{bmatrix} ({\bm{\mu}^{(0)}}^\top\bm x)^2-\tau^2-\lambda & \ ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x)& \ \dots & \ ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(M)}}^\top\bm x)& \ \dots & \ ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & \ ({\bm{\mu}^{(1)}}^\top\bm x)^2 & \ \dots & \ ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(M)}}^\top\bm x)& \ \dots & \ ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & \ ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x)& \ \dots & \ ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(M)}}^\top\bm x) & \ \dots & \ ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ \vdots & \ \vdots & \ \ddots & \ \vdots & \ \ddots & \ \vdots \\ ({\bm{\mu}^{(M)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & \ ({\bm{\mu}^{(M)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x)& \ \dots & \ ({\bm{\mu}^{(M)}}^\top\bm x)^2+\lambda & \ \dots & \ ({\bm{\mu}^{(M)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ \vdots & \ \vdots & \ \ddots & \ \vdots & \ \ddots & \ \vdots \\ ({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & \ ({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x)& \ \dots &\ ({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(M)}}^\top\bm x) & \ \dots & \ ({\bm{\mu}^{(n)}}^\top\bm x)^2\\ \end{bmatrix} \end{aligned}

is positive semidefinite, where M is the index of \sup_j \zeta_j^2.

Theorem

Theorem 0.4

6PW0

Consider the uncertain portfolio problem (?) under polyhedral uncertainty. The solution vector \bm x to the problem is feasible for the robust constraint in (?) if and only if there exists a \lambda \geq 0 such that the matrix

Scriptsize.
\begin{aligned} \begin{bmatrix} ({\bm{\mu}^{(0)}}^\top\bm x)^2-\tau^2-\lambda & ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x) & ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x) & \dots & ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(1)}}^\top\bm x)^2+\lambda & ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x)+\lambda & \dots & ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)+\lambda\\ ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x)+\lambda & ({\bm{\mu}^{(2)}}^\top\bm x)^2+\lambda & \dots & ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)+\lambda\\ \vdots & \vdots & \vdots & \ddots & \vdots \\ ({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x)+\lambda &({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x)+\lambda & \dots & ({\bm{\mu}^{(n)}}^\top\bm x)^2+\lambda \end{bmatrix} \end{aligned}

is positive semidefinite.

Theorem

Theorem 0.5

682F

Consider the uncertain portfolio problem (?) under the uncertainty set obtained from the combination of box and ellipsoidal uncertainty sets. Let the radii of the box and ellipsoidal sets are respectively \delta_B and \delta_E. Then the solution vector \bm x to this problem is feasible for the robust constraint in (?) if and only if there exists a \lambda \geq 0 such that the following matrix inequalities hold:

Tiny.
\begin{aligned} \begin{aligned} & \left \{ \begin{array}{ll} \begin{bmatrix} (\sqrt n{\bm{\mu}^{(0)}}^\top\bm x)^2-n \tau^2-n\lambda & (\sqrt n{\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x) & (\sqrt n{\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x) & \dots & (\sqrt n{\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ ({\bm{\mu}^{(1)}}^\top\bm x)(\sqrt n{\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(1)}}^\top\bm x)^2+\lambda & ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x) & \dots & ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ ({\bm{\mu}^{(2)}}^\top\bm x)(\sqrt n{\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x) & ({\bm{\mu}^{(2)}}^\top\bm x)^2+\lambda & \dots & ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ \vdots & \vdots & \vdots & \ddots & \vdots \\ ({\bm{\mu}^{(n)}}^\top\bm x)(\sqrt n{\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x) &({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x) & \dots & ({\bm{\mu}^{(n)}}^\top\bm x)^2+\lambda\\ \end{bmatrix} \succeq 0, &\mbox{ if } \min\left\{n\delta_B^2, \delta_E^2\right\}=n\delta_B^2\\ \\ \begin{bmatrix} ({\bm{\mu}^{(0)}}^\top\bm x)^2-\tau^2-\lambda & ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x) & ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x) & \dots & ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(1)}}^\top\bm x)^2+\lambda & ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x) & \dots & ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x)& ({\bm{\mu}^{(2)}}^\top\bm x)^2+\lambda & \dots & ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ \vdots & \vdots & \vdots & \ddots & \vdots \\ ({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x) &({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x) & \dots & ({\bm{\mu}^{(n)}}^\top\bm x)^2+\lambda \end{bmatrix} \succeq 0, &\mbox{ if } \min\left\{n\delta_B^2, \delta_E^2\right\}=\delta_E^2 \end{array} \right. \end{aligned} \end{aligned}
Theorem

Theorem 0.6

4EJJ

Consider the uncertain portfolio problem (?) under the uncertainty set obtained from the combination of box and polyhedral uncertainty sets. Let the radii of the box and polyhedral sets are respectively \delta_B and \delta_P. Then the solution vector \bm x to this problem is feasible for the robust constraint in (?) if and only if there exists a \lambda \geq 0 such that the following matrix inequalities hold:

Tiny.
\begin{aligned} \begin{aligned} & \left \{ \begin{array}{ll} \begin{bmatrix} (n{\bm{\mu}^{(0)}}^\top\bm x)^2-n^2\tau^2-n^2\lambda & (n{\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x) & (n{\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x) & \dots & (n{\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ ({\bm{\mu}^{(1)}}^\top\bm x)(n{\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(1)}}^\top\bm x)^2+\lambda & ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x)+\lambda & \dots & ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)+\lambda\\ ({\bm{\mu}^{(2)}}^\top\bm x)(n{\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x)+\lambda & ({\bm{\mu}^{(2)}}^\top\bm x)^2+\lambda & \dots & ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)+\lambda\\ \vdots & \vdots & \vdots & \ddots & \vdots \\ ({\bm{\mu}^{(n)}}^\top\bm x)(n{\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x)+\lambda &({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x)+\lambda & \dots & ({\bm{\mu}^{(n)}}^\top\bm x)^2+\lambda \end{bmatrix} \succeq 0, &\mbox{ if } \min\left\{n\delta_B, \delta_P\right\}=n\delta_B\\ \\ \begin{bmatrix} ({\bm{\mu}^{(0)}}^\top\bm x)^2-\tau^2-\lambda & ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x) & ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x) & \dots & ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(1)}}^\top\bm x)^2+\lambda & ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x)+\lambda & \dots & ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)+\lambda\\ ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x)+\lambda & ({\bm{\mu}^{(2)}}^\top\bm x)^2+\lambda & \dots & ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)+\lambda\\ \vdots & \vdots & \vdots & \ddots & \vdots \\ ({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x)+\lambda &({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x)+\lambda & \dots & ({\bm{\mu}^{(n)}}^\top\bm x)^2+\lambda\\ \end{bmatrix} \succeq 0, &\mbox{ if } \min\left\{n\delta_B, \delta_P\right\}=\delta_P \end{array} \right. \end{aligned} \end{aligned}
Theorem

Theorem 0.7

TRAC

Consider the uncertain portfolio problem (?) under the uncertainty set obtained from the combination of ellipsoidal and polyhedral uncertainty sets. Let the radii of the ellipsoidal and polyhedral sets are respectively \delta_E and \delta_P. Then the solution vector \bm x to this problem is feasible for the robust constraint in (?) if and only if there exists a \lambda \geq 0 such that the following matrix inequality hold:

\begin{aligned} \begin{scriptsize} \begin{bmatrix} ({\bm{\mu}^{(0)}}^\top\bm x)^2-\tau^2-\lambda & ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x) & ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x) & \dots & ({\bm{\mu}^{(0)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(1)}}^\top\bm x)^2+\lambda & ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x) & \dots & ({\bm{\mu}^{(1)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x)& ({\bm{\mu}^{(2)}}^\top\bm x)^2+\lambda & \dots & ({\bm{\mu}^{(2)}}^\top\bm x)({\bm{\mu}^{(n)}}^\top\bm x)\\ \vdots & \vdots & \vdots & \ddots & \vdots \\ ({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(0)}}^\top\bm x) & ({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(1)}}^\top\bm x) &({\bm{\mu}^{(n)}}^\top\bm x)({\bm{\mu}^{(2)}}^\top\bm x) & \dots & ({\bm{\mu}^{(n)}}^\top\bm x)^2+\lambda \end{bmatrix} \end{scriptsize} \succeq 0. \end{aligned}

Source and licence

The text on this page is from arXiv:2311.04609v1 by Pulak Swain and Akshay Kumar Ojha, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/2311/2311.04609.jsonl.