Sprachen

L3S Research Center -  RSS Feeds

Sie sind hier

Startseite » Veranstaltungen » Ausgewählte Themen der Informatik

Informatik Kolloquium: Aktuelle Themen der Informatik

Gemeinsam mit der Fakultät für Elektrotechnik und Informatik und der Regionalgruppe Hannover der Gesellschaft für Informatik organisiert das L3S im Wintersemester 2018 / 2019 ein Kolloquium zu aktuellen Themen der Informatik. Alle Vorträge finden in der Appelstrasse 9a im 15. Stock statt.

Dr. Markus Wulfmeier

Scalable and Efficient Robot Learning

Dr. Markus Wulfmeier, Google DeepMind / Oxford Robotics Institute

Automation and applications of robots in various fields bear the promise of reducing expenses as well as time requirements for production, logistics, transportation, and others. The first step towards automation included writing down our own rules and intuitions about how machines should solve tasks: programming. Machine learning enables us to generate rules which are too complex to be manually formulated by training highly flexible models based on large datasets. Our efforts have been shifted from rule design to the collection, cleaning, and annotation of data. To overcome increasing time demands for larger and larger datasets, we rely on methods from fields such as transfer learning, domain adaptation, learning from demonstration and reinforcement learning. In this talk, I will summarise some of our recent work from the Oxford Robotics Institute (University of Oxford) and the Berkeley AI Research lab (UC Berkeley) aiming at conceptualising the current challenges as well as the potentials for increasing the efficiency of humans to increase the efficiency of robotic automation.

9 November 2018, 14:00
MZ2, Erdgeschoss, Appelstr. 9A

Prof. Dr. Alice McHardy

Machine Learning in Infection Research

Prof. Dr. Alice McHardy, HZI Helmholtz Centre for Infection Research

The Department of “Computational Biology for Infection Research” studies the human microbiome, viral and bacterial pathogens, and human cell lineages within individual patients by analysis of large-scale biological and epidemiological data sets with computational techniques. Focusing on high throughput meta’omics, population genomic and single cell sequencing data, we produce testable hypotheses, such as sets of key sites or relevant genes associated with the presence of a disease, of antibiotic resistance or pathogenic evasion of immune defense. In my talk, I will give an overview of these topics and show with some examples how machine learning and data mining methods can play a big role in the analysis of these data, for a range of biological classification or regression tasks.

16 November 2018, 13:00
Multimedia Room, 15th Floor, L3S

Prof. Dr. Anoush Margaryan

Workplace learning practices in crowdwork platforms

Prof. Dr. Anoush Margaryan, Centre for Learning in the Platform Economy (LeaP), University of West London, UK

Crowdworkers are an online workforce operating outside organisational settings; they have no access to formal training and professional development opportunities that employees typically do (Kuek et al, 2015). Crowdwork has been criticised for being low in learning-intensity, causing deskilling and preventing workers from developing and applying their skills (Degryse, 2016). Emergent empirical findings challenge these accounts indicating that crowdworkers engage in self-initiated learning (Margaryan, 2016) and develop skills, such as business development, marketing, digital literacy and technical skills (Barnes et al, 2015; Gupta, 2017) through their work on the platforms. Yet how crowdworkers go about learning and professional development in the crowd workplace, what workplace learning activities and learning strategies they utilise to self-regulate their learning is not well understood. In this talk, findings from a series of recent surveys and interviews with crowdworkers, platform providers and other stakeholders will be presented to begin to elucidate workplace learning practices in crowdwork platforms. The talk is based on the following key ongoing research projects: 'Skill formation in online platform work' (in collaboration with Oxford Internet Institute) and  'Learning in Crowdwork'  (in collaboration with Frankfurt University)

23 November 2018, 14:00
Multimedia Room, 15th Floor, L3S

Prof. Dr. Andreas Bernard

Persons of Interest: The Status of the Self in Digital Cultures

Prof. Dr. Andreas Bernard, Leuphana University

What is striking about today’s methods of self-representation and self-perception – the profiles of social media but also the various locational functions on smartphones or the bodily measurements of the “quantified-self movement” – is the fact that they all derive from methods of criminology, psychology, or psychiatry that were conceived at various points since the end of the nineteenth century. Certain techniques for collecting data, which were long used exclusively by police detectives or scientific authorities to identify suspicious groups of people, are now being applied to everyone who uses a smartphone or social media. Biographical descriptions, GPS transmitters, and measuring devises installed on bodies are no longer just instruments for tracking suspected criminals or patients but are now being used for the sake of having fun, communicating, making money, or finding a romantic partner. The talk tries to trace back these genealogies and thus present the criminological fundaments of contemporary subjectivity.

11 Januar 2019, 14:00
Multimedia Room, 15th Floor, L3S

Prof. Dr. Paolo Boldi

A Modern View of Centrality Measures

Prof. Dr. Paolo Boldi, Università degli Studi di Milano

Given a social network, which of its nodes are more central? This question was asked many times in sociology, psychology and computer science, and a whole plethora of _centrality measures_ were proposed to account for the importance of the nodes of a network.
Also, many modern IR problems call for a mixture of classical text-retrieval methods with graph mining techniques, especially within the area of social-network search, where node centrality can play a crucial role.

But what do existing centrality measures actually measure? To what extent do they agree? What is their meaning when they disagree? And further: which of them can be computed or approximated reasonably on the large (huge) graphs under observation?

My talk is twofold. On one hand, I will try to give a comprehensive and mathematically coeherent account of the most important centrality measures from the literature, and present for the first time a set of scientifically
well-grounded methodologies to establish whether a measure is actually doing what it has been designed for. I propose to assume an axiomatic approach, wherein I suggest some simple, basic properties a centrality measure should have, and I corroborate my findings with some anectodal results on publicly available medium-to-large web and social networks. I validate the results by examining each measure under the lens of information retrieval, leveraging state-of-the-art knowledge in the discipline to measure the effectiveness of the various indices in locating web pages that are relevant to a query.

On the other hand, I approach the problem of computing a large family of centralities (known as "geometric centralities") on very large graphs accessed in a semi-streaming fashion, with a very small amount of memory (in the order of a dozen bytes) per node available in core memory. I leverage the newly discovered algorithms based on HyperLogLog counters, making it possible to approximate a number of geometric centralities at a very high speed and with high accuracy. While the application of similar algorithms for the approximation of some measures (e.g., closeness) is not new, our exploitation of HyperLogLog counters reduces exponentially the memory footprint, paving the way for in-core processing of networks with a hundred billion nodes using "just" 2 TiB of RAM. Moreover, the computations I describe are inherently parallelizable, and scale linearly with the number of available cores.

8 Februar 2019, 14:00
Multimedia Room, 15th Floor, L3S