PhD Seminar: Summer term 2009
(Budapest University of Technology and Economics, Chair: Denes Bisztray, Host: Reiko Heckel)
Graph transformation provides a high-level rule and pattern-based manipulation language for graph models. In graph transformation, the most cost-intensive phase of a transformation execution is pattern matching, where those subgraphs of a model graph are identi?ed and matched which satisfy constraints prescribed by graph patterns. Incremental pattern matching aims to improve the ef?ciency of this critical step by storing the set of matches of a graph transformation rule and incrementally maintaining it as the model changes, thus eliminating the need of recalculating existing matches of a pattern. In recent research papers, we have shown through benchmark examples that incremental pattern matching has a significant advantage in the application domain of model simulation and model synchronization. In the presentation, I describe our RETE-based incremental pattern matcher implementation included in the VIATRA2 model transformation framework. Moreover, through simple examples, I show basic techniques of combining local search-based pattern matching with the incremental technology to further increase performance.
(ESA, Chair: Christian Kissig, Host: Emilio Tuosto)
The presentation consists of two parts. Firstly, I present the result of my master thesis "On Secure Services for Embedded Mobile Peer-to-peer Systems: a Conceptual Framework and its Implementation in the Coordination Language SecureLime". In the context of the European Project SMEPP, the thesis describes the implementation of a middleware built on top of a Linda-like language. Then, I outline a possible research topic for a PhD thesis which would be partly based on my master thesis. Finally, I present quickly my current work concerning the simulation of communication protocols for Mars exploration.
(University of Birmingham, Host: Daniela Petrisan)
I present several ways of representing (a subset of) the real numbers and outline why none of them is fully satisfying yet. A unifying view on these representations is given which leads to the question whether a nice representation can exist at all. I present a bitopological operator which has the real numbers as a fixed point.
(Imperial College, London, Host: Christian Kissig)
Scenario-based speci?cations are popular ways for describing intended system behaviour. With our research we aim to facilitate early analysis of system behaviour and the development of behaviour models in conjunction with scenarios.
In this talk I will present a scenario formalism with trigger and branching flavoured semantics. This formal language is consistent with existing informal scenario-based and use-case based approaches to requirements engineering. I will also show a synthesis algorithm that, rather than building arbitrarily one of the many behaviour models that satisfies a scenario, constructs a Modal Transition System (MTS) which characterizes all behaviour models that conform to the scenario. The triggered scenarios and the synthesis algorithm have been included in the tool MTSA (Modal Transition System Analyser).
I will conclude the presentation with a very brief demo of the tool.
Eylem Guezel Karpuz
(Balikesir University, Turkey, Chair: Daniela Petrisan, Host: Rick Thomas)
The property of having finite derivation type, FDT for short, is a finiteness condition for finitely presented monoids. It is obtained by considering certain binary relations, called homotopy relations, on the set of paths of a graph that is associated with a finite presentation [Sigma; S] of the monoid M considered. This property was introduced by C. Squier. Since FDT is a necessary condition for a monoid to be defined by some finite and complete string- rewriting system, and since FDT is an invariant property of monoid presentations, it becomes important to know which monoid constructions preserve the property of FDT.
In this talk, I will summarize some works studied on FDT and then give a brief outline of my study on FDT.
(Kings College, London, Host: Christian Kissig)
Protein folding is the physical process in which a protein acquires its three-dimensional structure, which subsequently determines the specific biological task performed by this protein. Proteins fold spontaneously into three-dimensional structures of minimum energy when found in the appropriate environment. Moreover the sequence composition of a protein is sufficient to determine its structure. We deal with the ab-initio protein structure prediction problem where, starting from the sequence composition, the aim is to find the global energy minimum in a prescribed model of interactions. This is an NP-complete problem in the simplest coarse grained model, namely the HP model. In order to tackle the problem we devise a novel stochastic local search method, the Population Based Local Search. Our method employs simulated annealing as a pre-processing step, to estimate the maximum over the minimum escape height from any local minimum. This information is used to guide the Population-based local search. The method is based on independent guided random walks which sample local minima. The sampled local minima can subsequently be used for a statistical analysis of the underlying energy landscape. The novel method competes well to a pure simulated annealing and also has the potential of an efficient parallel implementation.
(University of Nottingham, Chair: Ambreen Shahnaz, Host: Martin Birks)
This presentation introduces the research areas related to wireless ad hoc networking, identifying two classifications of mobile wireless user - One who is typically connected, and another that is typically disconnected. I then show that a modern mobile device can be connected or disconnected, and as such cannot be classified as either a typically connected or disconnected node, and can be best described as a node of varying connectivity.
I present a routing protocol built upon the assumption that mobile users have a reoccurring, predictable set of contacts. The protocol I propose exploits the assumed connectivity knowledge, resulting in a more efficient approach to packet transferal.
I finish the presentation describing the results I have obtained from my emulations in the network simulator ns-2.
(University of Essex, Host: Ambreen Shahnaz)
A taxonomy of approaches could be developed for the work that has been carried out with respect to topology generation for representative Internet and router level Topologies.
We suggest that there are fundamentally two different requirements for accurate Internet topology generation. The fist relates to the AS-level. There are existing and well developed AS long-tailed distribution models, which are clearly suitable and well accepted for the AS-level topologies. Many of these methods use the placement of nodes representing AS (organizations) on an xy plane. These AS's are then connected into a graph using mathematical formula proven to result in a graph similar to the internet connectivity, Unfortunately the formulas used to connect the nodes (representing AS's) assumes link independency. Analysis of these models (Barabasi, Waxman, GLP), has shown the resulting graphs to represent the Internet AS structure quite well. However we propose that for router level topologies this independence is not present and suggest an over-riding distance/price heuristics model for Internet/ Corporate router level topology generation.
We suggest that the AS-level methods when applied are in fact antithetical for router level topology generation; ie, they create topologies that can represent the worst graph. Although many existing topology generators already have 'router level' settings, the models used to create /evaluate these router topologies has remained with some exceptions based on the power-law distributions.
(University of Copenhagen; Chair: Laura Bocchi)
This talk reveals the intention and results of extending the Multiparty Session Types with Symmetric Sum. A Clinical Practice Guideline (CPG) is a detailed description of best practice for a medical treatment. First the desired properties of a representation of CPGs are discussed. Then a specific representation called Process Matrix is described. Symmetric Sum Types are introduced as a way to model Process Matrix as Multiparty Session Types. Finally the theoretic properties and results of Symmetric Sum Types such as encodability are presented.