MEMICS
MEMICS
Doctoral Workshop on Mathematical and
Engineering Methods in Computer Science
 
organized jointly by the Masaryk University
and the Brno University of Technology, Czechia
2015
NEWS
 
May 9, 2016
MEMICS 2016
We invite you to Telč for MEMICS 2016 conference.
November 9, 2015
Best Papers
Conference proceedings and photogallery are available. Best Papers announced.
MAIN FACTS
MEMICS IN BRIEF
INVOLVEMENT
IMPORTANT DATES
PROGRAMME
INVITED SPEAKERS
PAPERS
PRESENTATIONS
PASTIME
SOCIAL EVENTS
LOCALE
VENUE
MAP
October 23—25, 2015    University Centre    Telc    Czech Republic
 

Invited speakers
Ezio Bartocci
Vienna University of Technology, Austria
Formal Methods for Monitoring and Synthesis of Spatio-Temporal Properties
Networked dynamical systems are increasingly used as models for a variety of processes ranging from robotic teams to collections of genetically engineered living cells. As the complexity of these systems increases, so does the range of emergent properties that they exhibit. In pattern recognition and machine learning, spatio-temporal properties are usually classified according statistical descriptors (or features). This approach despite its success and popularity lacks of a rigourous foundation to specify such patterns and to reason about them in a systematic way. On the other end, formal methods provide logic-based languages with a well-defined syntax and semantics to specify in a precise and concise way emergent behaviours and the necessary techniques to automatically detect them. Formal analysis is however challenging due to the difficulty of defining spatio-temporal properties in a suitable formal language. In this talk. we will present some recent results on novel formal logic-based languages and methods to monitor and synthetize spatio-temporal properties in networked dynamical systems.

Siegfried Benkner
Faculty of Computer Science, University of Vienna
Programming Support for Future Parallel Architectures
Due to physical constraints the performance of single processors has reached its limits, and all major hardware vendors switched to multi-core architectures. In addition, there is a trend towards heterogeneous parallel systems comprised of conventional multi-core CPUs, GPUs, and other types of accelerators, mainly because such system promise a better performance/power ratio. As a consequence, the development of applications that can exploit the full potential of emerging parallel systems is becoming more and more challenging. In this talk we outline the major challenges associated with software development for future parallel architectures, including higher-level parallel programming models, advanced runtime technology, and automatic performance tuning tools. In particular we will present a high-level compositional approach to parallel software development in concert with an intelligent runtime system and automatic performance tuning techniques. Such an approach can significantly enhance programmability of future parallel systems, while ensuring efficiency and performance portability across a range of different architectures. We report on recent research results of two European projects, PEPPHER and AUTOTUNE, which addressed the challenges of software development for current and emerging parallel systems, and discuss related research efforts and potential future directions.

Mike Just
Heriot-Watt University, Edinburgh, Scotland
Recent trends in authentication research
The past five years have been especially busy for the study of authentication techniques, such as passwords. In this talk I will review several recent techniques for improving authentication security and usability, including the use of honey passwords and renewed investigations into the use of randomly generated passwords. In addition, I will discuss the evolution of metrics that are used for measuring password security, including entropy and more recent attempts to model the guessing patterns of attackers.

Simone Severini
University College London
Graphs and quanta
I will talk about various ways to associate a graph to a quantum mechanical state or a quantum operation. This is an old story but still a good excuse to use quantum theory to play with graphs and graphs to play with quantum theory.

Natasha Sharygina
University of Lugano, Switzerland
Flexible interpolation for efficient model checking
Craig interpolation is widely used in symbolic model checking. Most interpolation algorithms construct the interpolant from a proof of unsatisfiability and a fixed labeling determined by which part of the propositional formula is being over-approximated. While this results in interpolants with fixed strength, it limits the possibility of generating interpolants of small size. This is problematic since the interpolant size is a determining factor in achieving good overall performance in model checking. This talk presents a new framework with a flexible labeling approach to construct small interpolants. In addition to providing the labeling mechanism guaranteeing small interpolants, this framework is capable to manage the strength of the interpolants. This talk presents the flexible interpolation mechanism in action by demonstrating how it works in incremental model checking. We will discuss two different applications: verification of multiple properties of the same code and verification of software upgrades. Both use interpolation as a backbone technology and face different strength requirements. We will demonstrate how flexible interpolation outperforms any of the fixed interpolation algorithms.

Peter Vojtáš
Charles University, Czech Republic
Understanding transparent and complicated users as instances of preference learning for recommender systems
Authors: P. Vojtas, M. Kopecky, M. Vomlelova
In this talk we are concerned with user understanding in content based recommendation. We assume we have a user-item matrix with explicit rating and a time stamp. Our task is to learn/predict user preference on unseen/unrated items.
In this talk we are not going to present new mining technics nor improve achievements of comparable efforts on some data and metrics. Rather, we concentrate on three issues:
First, we study three different data sets from the same domain of movie recommendation, trying to avoid features specific for single data and try to be more generic. We consider also semantic enrichment of movie data to enable content based recommendation.
Second, we use several metrics which were not used so far in the recommender systems domain. Besides classical rating approximation with RMSE and top-k we study new metrics for predicting next-k and (at least) k-hit. We consider those parameters k variable and try to measure their behavior
Finally and most importantly, we trace performance of our methods and metrics as distribution along each single user. This helps us to detect transparent and complicated users. Improving these metrics (for instance 1-hit) can contribute to the increase of user satisfaction.
We provide results of experiments with several combination of methods, data sets and metrics along of these three axes. Nevertheless, all our experiments are offline on public accessible data. A validation in real online A/B testing needs access to an application. We were not able to realize this so far. We concentrate only on algorithmic part of recommendation, business and marketing part of recommendation problem is out of scope of this study.

 
 
 
MEMICS
 
Copyright © FI MU
Copyright © FIT VUT
Brno, 2005–2014
FIFIT
Photos and Design