## CWI Scientific Meeting

Dear colleagues,

This is the second announcement of the CWI Scientific Meeting that will be held this Friday, June 1. We'll have six presentations by PhD students from throughout the institute. As usual, sandwiches will be provided before the talks.

We hope to see you at the meeting!

The organizers,

Willem Hundsdorfer, Ronald de Wolf

PROGRAM

Date: Friday, June 1

Time: 13.00 - 14.00

Room: Turing (NB: the big room, not the usual smaller one)

Speakers and titles:

Bart de Keijzer (PNA1): *Algorithmic Game Theory*

Joost Bosman (PNA2): *Extreme value theory in a tandem of fluid queues*

Willem Haverkort (MAC2): *Magnetohydrodynamic stability of rotating tokamak plasmas*

Inken Wohlers (MAC4): *Exact algorithms for protein structure alignment*

Jeroen van den Bos (SEN1): *Model-Driven Engineering in Digital Forensics*

Holger Pirk (INS1): *Data Management in the Age of Heterogeneous Hardware*

Abstracts:

Bart de Keijzer (PNA1): Algorithmic Game Theory

Games are situations in which multiple self-interested participants (called "players") compete against each other and have to make competitive decisions. Game theory is the mathematical study of games. Algorithmic game theory is the subfield that studies the computational questions related to game theory. In this talk I give an extremely quick introduction to algorithmic game theory. Subsequently, I move on to algorithmic questions that can be posed in case the players behave cooperatively or are not completely selfish (as often classically assumed). I illustrate this by giving some examples from my research and stating (informally) some of the research results associated to these examples.

Joost Bosman (PNA2): Extreme value theory in a tandem of fluid queues

We consider a tandem of two fluid queues. The first fluid queue is a Markov Modulated fluid queue and the second queue has a fixed output rate. We show that the distribution of the total fluid converges to a Gumbel (extreme value) distribution. Using this result we approximate the probability that the second buffer becomes empty. This model is motivated by a video streaming application over an unreliable communication network. The results can be used to dynamically control video quality under uncertain and time-varying network conditions.

Willem Haverkort (MAC2): Magnetohydrodynamic stability of rotating tokamak plasmas

The longstanding dream of achieving controlled nuclear fusion for the production of energy draws closer with the construction of the ITER tokamak in France. In a tokamak, a plasma hotter than the core of the sun floats in a magnetic field without touching the walls. In present tokamaks the plasma rotates, which can significantly influence stability. This can be described quite well with the theory of magnetohydrodynamics. Using various simplifications, we derived an insightful stability criterion that reveals the various ways in which a rotating plasma can become unstable.

Inken Wohlers (MAC4): Exact algorithms for protein structure alignment

A protein structure alignment is a sequential one-to-one mapping between structurally equivalent residues in two proteins. In my talk, I will illustrate the protein structure alignment problem and its importance and motivate why and how we use exact algorithms to solve it.

Jeroen van den Bos (SEN1): Model-Driven Engineering in Digital Forensics

Software tools used in digital forensics investigations are difficult to develop and maintain. The constant increase in new devices and applications means forensics software must be continually adapted and extended. The constant increase in storage capacity and bandwidth means forensics software must be very fast and scalable. In our research we apply model-driven engineering in an attempt to make these requirements easier to achieve and maintain.

Holger Pirk (INS1): Data Management in the Age of Heterogeneous Hardware

Computer Systems don't just become increasingly parallel, they become increasingly diverse as well. Traditional harddisks can be assisted by solid-state disks, main memory by hierarchical caches and conventional CPUs by GPGPUs. While this wealth of available hardware holds great potential for many custom-made applications, the efficient use in generic Database Management Systems is still an open problem. We present our research on the efficient co-operation of heterogeneous hardware components for large scale data management.