1. Speaker: Prof. Sergiy Butenko, Texas A&M University, USA

http://ise.tamu.edu/people/faculty/butenko/

Title: The Maximum Independent Union of Cliques Problem: Complexity and Exact Approaches

Co-authors: Zeynep Ertem, Eugene Lykhovyd, Yiming Wang, and Sergiy Butenko

Abstract: Given a simple graph, the maximum independent union of cliques problem is to find a maximum-cardinality subset of vertices such that each connected component of the corresponding induced subgraph is a complete graph. This recently introduced problem allows both cliques and independent sets as feasible solutions and is of significant theoretical and applied interest. This paper establishes the complexity of the problem on several classes of graphs (including planar, claw-free, and bipartite graphs), and develops an integer programming formulation and an exact combinatorial branch-and-bound algorithm for solving it. Results of numerical experiments with numerous benchmark instances are also reported.

2. Speaker: Corresponding member of NAS of Kazakhstan, Prof. Maksat Kalimoldayev, IICT, Kazakhstan

Title: Mathematical and Computer Modeling of Stability of Complex Electric Power Systems

Abstract: This article discusses the development and study of mathematical model of complex power systems for the global asymptotic stability problems. The conditions have been obtained for the global asymptotic stability of nonlinear control systems. Control actions that ensure stabilization of complex electric power systems have been found. The software package of dynamic study of complex electric power systems has been developed in Visual Studio using the programming language C-Sharp.

3. Speaker: Prof. Yurii Nesterov, CORE/UCL, Belgium, Higher School of Economics, Russia

https://uclouvain.be/fr/repertoires/yurii.nesterov

https://www.hse.ru/org/persons/150293981

Title: Accelerating the Universal Newton Methods

Abstract: In this talk we present new results related to the universal second-order methods, which can automatically adjust the level of smoothness of the objective function. Our methods converge in accordance to the best rates allowed by a Holder condition introduced for the Hessians. As compared with the usual Newton method, the reason for acceleration of our schemes consists in accumulation of global information on the behavior of the objective, represented by a linear model. Our methods solve the convex optimization problems in composite form.