update math display (abstracts)

Author: sebcolla

february_2023/program.html

@@ -205,20 +205,20 @@ <h2>Monday the 13th</h2>
                     with Linear Mappings  </strong> </summary>
                     We develop a Performance Estimation Problem representation for linear mappings. We consider convex
                     optimization problems involving linear mappings, such as those whose objective functions include
-                    compositions of the type g(Mx), or featuring linear constraints of the form Mx=b. First-order methods
+                    compositions of the type \(g(Mx)\), or featuring linear constraints of the form \(Mx=b\). First-order methods
                     designed to tackle these problems will typically exploit their specific structure and will need to
-                    compute at each iteration products of iterates by matrices M or M^T.
+                    compute at each iteration products of iterates by matrices \(M\) or \(M^T\).
                     <br>
                     Our goal is to identify the worst-case behavior of such first-order methods, based on the Performance
                     Estimation Problem (PEP) methodology. We develop interpolation conditions for linear operators M and
-                    M^T. It allows us to embed them in the PEP framework, and thus, to evaluate the worst-case performance
+                    \(M^T\). It allows us to embed them in the PEP framework, and thus, to evaluate the worst-case performance
                     of a wide variety of problems involving linear mappings. We cover both the symmetric and nonsymmetric
                     cases and allow bounds on the spectrum of these operators (lower and upper bounds on the eigenvalues
                     in the symmetric case, maximum singular value in the nonsymmetric case). As a byproduct we also
                     obtain interpolation conditions and worst-case performance for the class of convex quadratic functions.
                     <br>
                     We demonstrate the scope of our tool by computing several tight worst-case convergence rates,
-                    including that of the gradient method applied to the minimization of g(Mx) and that of the
+                    including that of the gradient method applied to the minimization of \(g(Mx)\) and that of the
                     Chambolle-Pock algorithm. <br> <br>
                 </details>