• de

19.10.2016 - no meeting

26.10.2016 - Francesco Quinzan

EAs in Real-World Optimization: Noise, Performance, Applications

When dealing with optimization of real-world numerical models, it is often unfeasible to perform correlation via direct methods. This is the case of the thermal analysis of bespoke electro-mechanical systems for extreme harsh environments, such as actuators and wireless sensors. Evolutionary Algorithms (EAs) are generic population-based optimization algorithms, useful to approach such problems.

Even though in the last decade theoretical results have been presented regarding the running time of EAs, there is still little understanding of the interplay between the operators of randomized search heuristics and explicit noise-handling techniques, such as statistical resampling. We report on several statistical models and theoretical results that help to clarify this reciprocal relationship for a collection of randomized search heuristics on noisy functions.

We consider the optimization of pseudo-Boolean functions under additive posterior Gaussian noise and explore the trade-off between noise reduction and the computational cost of resampling. We perform experiments to find the optimal parameters at a given noise intensity for a mutation-only evolutionary algorithm, a genetic algorithm employing recombination, an estimation of distribution algorithm (EDA), and an ant colony optimization algorithm. We observe how the optimal parameter depends on the noise intensity for the different algorithms. We locate the point where statistical resampling costs more than it is worth in terms of run time.

We find that the EA requires the highest number of resamples to obtain the best speed-up, whereas crossover reduces both the run time and the number of resamples required. We find that EDA-like algorithms require no resampling, and can handle noise implicitly. Interestingly, we observe that all tested algorithms exhibit high variance for increasing noise. We plan to combine the latter observation with rigorous analysis to define the optimal re-start strategy, with the hope of further improving performance.

02.11.2016 - Nanjing Feedback Round

09.11.2016 - Anton Krohmer

Structures & Algorithms in Hyperbolic Random Graphs

Complex networks like social networks, biological networks or computer networks are ubiquitous in nature and society. In this talk, we introduce a generative mathematical model to study such networks, called „Hyperbolic Random Graphs“. We theoretically analyze their structural properties, for instance the emergence of cliques and its diameter. Furthermore, we present an algorithm for embedding real-world networks into the hyperbolic plane such that connected nodes are nearby, and disconnected nodes are placed far apart. We perform an extensive experimental evaluation to ensure its quality, and also apply it to a real-world graph as a proof of concept.

16.11.2016 - Kateryna Kuksenok

When Good Intentions Meet Reality: Qualitative Study of Programming Practices

Programming is undertaken by people many backgrounds for many purposes. This talk focuses on qualitative methods for the study of changing programming practices. Based on ethnographic observation and interviews, we developed a theory of deliberate individual change. Under this theory, a shared imagination of a perfect world drives prioritization and decision-making. It is an unattainable moving target against which current action is evaluated. The tension between the imagined and actual reality is explored both through an ethnography of oceanographers, and more recently in mining code repositories and revision histories. 

23.11.2016 - Sankalita Mandal

Re-evaluation of Decisions based on Events

The effective and efficient design and execution of business processes is a key driver for organizations. Therefore, organization capture their business processes in form of process models, with the industry-standard BPMN (Business Process Model and Notation) serving as guidance for the implementation of information systems. Recently, the importance of decision in business processes and structurally capturing them has been revived by the new Decision Model and Notation (DMN) standard. Decisions have a direct influence on the quality and performance of business processes, because they determine the course of business processes. Usually, when a decision is taken, a specific path is chosen immediately based on that or the output is used by the process later on. However, in the digital world, organizations have the possibility to access an abundance of data in real-time. This offers the possibility to re-evaluate decision. . In the current work, a re-evaluation technique for decisions is presented by using event processing methods. If relevant event arrives within a certain time frame, then it may change previously taken decision to improve the process outcome.

30.11.2016 - Erik Scharwächter

Contextual Change Detection in Large Sensor Collections

The monitoring of urban, societal and environmental variables like air pollution, mobile phone usage, or tropical vegetation cover over time through various sensors allows to detect interesting, unusual changes relevant for decision makers. Recently, novel approaches for contextual time series change (CTC) detection have been proposed that define changes within a time series with respect to the behavior of other, similar time series. We address two shortcomings of the existing approaches using remote sensing imagery as an example use case: (1) sensors deliver high-dimensional time-series with context and changes hidden in subspaces, but existing work is limited to the univariate case; and (2) context construction is the major bottleneck of existing approaches and does not scale for large numbers of time series.

07.12.2016 - Heng Gao

14.12.2016 - Ralf Rothenberger

21.12.2016 - Christmas Break

28.12.2016 - Christmas Break

04.01.2017 - Martin Krejca

11.01.2017 - to be announced

18.01.2017 - to be announced

25.01.2017 - to be announced

01.02.2017 - to be announced

08.02.2017 - to be announced