FORSKNINGSPRESENTATIONER
Andreas Lundell, Post doc.
I´m currently working on global optimization of MINLP problems containing signomial functions. This is an important class of problems since they are quite general, and appear in many different types of applications. However, since signomial functions are generally nonconvex, these problems are difficult to solve.
Henrik Nyman, PhD student
My research pertains to creating and improving algorithms that find the undirected graphical model which best represents the dependence structure of a multidimensional probability distribution. Such representations combine probability theory and graph theory. The research also aims to improve algorithms for data clustering, and for each data cluster find the graphical model that best represents the dependence structure within the given cluster.
Mikael Nyberg, PhD student
One of the research areas in the OSE group is to apply optimization methods on large scale industrial production planning problems. The goal is to create accurate and effective optimization models for various industries in order to improve productivity and minimize costs throughout the supply chain.
Ray Pörn, Post doc.
The Coulomb glass is a model for a disordered material, where the conduction electrons are localized to impurity sites and the electrons strongly interact with each other [1,2]. Given n possible sites where the k electrons can be localized, the problem is to find the electron configuration that minimizes the total energy of the system. A certain configuration of electrons in a material gives rise to a specific total energy of the system.
Mikael Kurula, Post doc.
I study infinite-dimensional linear systems from an input/output free perspective, mainly using the general but still well-structured class of passive state/signal systems. My main interests are the basic properties, realization theory and (rather general) interconnection theory of continuous-time state/signal systems. On a longer term, I am also interested in extending the state/signal theory to non-linear systems where this is feasible..
Anders Skjäl, PhD student
My work focuses on global optimization in nonconvex mathematical programming (NLPs, MINLPs). Algorithms for such problems usually include an element of branch-and-bound. I study various convex relaxations and the quality of the lower bounds they generate. I am particularly interested in twice differentiable constraint functions. The first and second order derivatives (gradient and Hessian) give enough information to construct convex underestimators of such expressions.
Toni Lastusilta, Post doc.
In research and industry the need to solve complex problems is evident and the use of optimization often provides a solution. A researcher might gain insights by solving a problem while an industry might find more efficient operational solutions. Therefore, it is important that the development in the optimization field is made accessible to its practitioners. The advances in optimization consist of theoretical findings that improve methods and models, as well as, the increase in computing power. Optimization platforms like AIMMS, AMPL and GAMS provide the interface to the optimization methods.
Axel Nyberg, PhD student
I’m studying the Quadratic Assignment Problem (QAP). Many real world problems can be formulated as quadratic assignment problems. These problems arise in economics, archeology, electronics, facility layout and many more different fields of research. Some real-life examples:
-Backboard Wiring...
-Hospital Design...
-Gate Assignment...
Amir Shirdel, PhD student
My research focuses on using support vector machines for identification of a dynamical system from experimental data, which forms an important problem in various control and signal processing tasks. Support vector Regression is a promising linear and nonlinear modeling method that has been found to perform very well in many fields, and has a powerful potential to be applied in system identification.
An important part of my research consists of developing identification methods for hybrid and switching systems, for which efficient general identification methods are largely lacking, especially for nonlinear systems.
Ralf Östermark, Professor
My research focuses on designing nonlinear optimization tools and vector-valued time series algorithms on parallel computers. The computational resources can be linked as support libraries to the Genetic Hybrid Algorithm (GHA). I have been developing GHA since the beginning of 2000. I have shown that the complexity of binary mixed-integer-nonlinear problems can be significantly reduced on parallel processors using asynchronic mesh interrupts and binary coding of local box constraints.
Ville-Pekka Eronen, PhD student
I am generalizing the ECP method and it's variants to cover nonsmooth problems. The work consists of creating appropriate modifications to the algorithm and proving convergence of the resulting algorithm.
Otto Nissfolk, PhD student
My research focuses on solving binary quadratic optimization problems. Binary quadratic problems arise for example in theoretical physics when looking into disordered materials. The Coulomb glass is a model for a disordered material, where the electrons are localized to impurity sites and the electrons strongly interact with each other. We try to find the ground state of these materials, that is the electron configuration with the lowest total energy.
Christer Glader, Post doc.
My research interests are in the area of approximation theory in the complex unit disk. For example, finite Blaschke products in connection with constructive methods for rational Chebyshev approximation, and various types of rational interpolation problems, so-called Nevanlinna-Pick interpolation, in the unit disk. Applications of such methods can be found in H∞-optimization and control theory (model reduction, model matching theory etc.).
Fredrik Jansson, Post doc.
One problem from my field of physics can be formulated as an optimization problem - finding the ground state of a Coulomb glass. The Coulomb glass ground state can then be studied with methods from the field of global optimization.
Johan Pensar, PhD student
My research pertains to developing the theory and designing inference algorithms for a class of probabilistic models called context specific graphical models. A context specific graphical model can represent a more refined dependence structure than an ordinary graphical model which can be considered a special case in this new class of models. Due to their enhanced flexibility, context specific models have the ability to represent more complex dependence structures.
Lassi Hietarinta, PhD student
I study modeling and identification of periodically time-varying dynamical systems. Such systems exhibit a significantly more complicated dynamic behaviour than non-periodic systems. Identification of sufficiently accurate models and controlling of such systems is therefore a non-trivial problem. The target of my study is to develop practically useful methods for identification and controlling of periodic systems.
John Eric Saxen, PhD student
Within intelligent control of internal combustion engines, there are still several aspects related to the system dynamics that need to be considered for better modelling and estimation. One such topic of interest in this research is multi-engine balancing, where a number of engine-generator sets are interconnected to a small-sized grid, which results in load fluctuations due to interactions between the engines.
OSE FORSKNING
The research in the OSE group is made up of two main blocks: Theory and methods, as well as Applications in science and engineering. The work within the blocks is divided into seven work packages.
Theory and methods
In this block we are interested in finding theoretical results that can be used in any of the implementations in the other block Applications and engineering. The group Theory and methods consists of the work packages: Optimization, Systems engineering, Systems theory and Mathematical statistics.
In Optimization the focus of the research is mixed integer global optimization, model representation and generic methods and techniques for global optimization of certain classes of large scale mixed-integer nonlinear problems.
In Systems engineering, one focus is on solving complex control problems, for which traditional controller design methods are inadequate. These problems include feedback control of systems which involve integer-valued variables, nonlinear systems and multiple objective controller design problems.
In Systems theory we look at the theory for controlling infinite-dimensional distributed systems. A natural setting in the case of a continuous time variable is to view them as infinite-dimensional well-posed linear systems. Another approach we are investigating is to replace the classical input/state/output point-of-view by a “state/signal” one.
In Mathematical statistics we study the Bayesian approach to statistics. Despite of the evident success of the Bayesian theory, there are also many research problems where the computational challenges have so far proven to be too exhaustive to promote wide-spread use of the state-of-the-art Bayesian methodology. To try to resolve this issue, we focus on stochastic computational and modeling strategies to develop algorithms that overcome problems associated with the analysis of highly complex data sets.
Applications in science and engineering
In this block, the theoretical work in Theory and methods is applied to a wide range of applications. The corresponding work packages in this group are Physics and material science, Biotechnology and –medicine, and Engineering.
In Physics and material science, we especially look at semi-conductor design in material physics through the means of optimization of electron configuration in disordered materials. This is in cooperation with the Physics department at Åbo Akademi University.
In Biotechnology and –medicine, we focus on the study of the global optimization part of so-called peptide docking problems. This type of problem can be found, e.g. in computational drug design. We also study mathematical modeling of complex dynamical cellular processes, such as gene regulatory networks. Statistical methods in medicine is another application we are investigating.
In Engineering we are studying large scale industrial scheduling, where complex production networks, especially in the manufacturing industry, are considered. Another field of study is time-periodic systems used, for instance, in active cylinder balancing methods for diesel engines.