Fakultätskolloquium der Technischen Fakultät im SS 2017

• Die Kolloquien an den Instituten für Elektrotechnik und Informationstechnik sowie Materialwissenschaften:

Jeweils montags eine halbe Stunde vor Beginn der Vorträge, also um 16.45 Uhr, werden die Gastredner den Besuchern des Kolloquiums in einer zwanglosen Runde bei einer Tasse Kaffee in der "Eisdiele" (Eingangshalle Geb. D, neben dem "Aquarium") vorgestellt.

Ansprechpartner:  Daniel Johannsen (Tel. 0431-880-6068) dekanat@tf.uni-kiel.de

• Die Kolloquien am Institut für Informatik:

Alle Vorträge finden - soweit nicht anders angegeben - im Raum Ü2/K des Instituts für Informatik (Ludewig-Meyn-Str. 2, Vorbau des Mathematischen Seminars) freitags um 14 Uhr c.t. statt. 45 Minuten vor Vortragsbeginn stehen Tee und Kaffee bereit.

Ansprechpartner/innen: Geschäftszimmer des Instituts für Informatik (Tel. 0431 880-7270)aktuell@informatik.uni-kiel.de

• Wenn Sie die Einladungen per E-Mail erhalten möchten tragen Sie sich bitte in die Mailingliste ein.

Bitte achten Sie bei Sonderkolloquien auf die Zeit- und Raumangaben!

Nachfolgend alle Vortragstermine in chronologischer Abfolge

Kolloquiumsvortrag, nn / am 03.04.2017

03.04.2017 von 17:15 bis 18:45

Institute Ostufer, Geb. D, "Aquarium", Kaiserstr. 2, 24143 Kiel

Kolloquiumsvortrag, nn / am 07.04.2017

07.04.2017 von 14:15 bis 15:45

Institut für Informatik, Ludewig-Meyn-Straße 2, 24118 Kiel, Raum: Übungsraum 2/K

Kolloquiumsvortrag, Dipl.-Ing. Ingo Schalk-Schupp, Forschungslabor Nuance Communication / am 10.04.2017

10.04.2017 von 17:15 bis 18:45

Institute Ostufer, Geb. D, "Aquarium", Kaiserstr. 2, 24143 Kiel

Titel: Nonlinear Echo Suppression

Abstract: This presentation provides a short overview concerning acoustic echo cancellation and acoustic echo suppression methods followed by a more in-depth discussion of new methods dealing with Hammerstein-type nonlinear distortions.

The Hammerstein system is divided into a linear and a parallel nonlinear part by an alternative way to define of the linear one. The implications of this separation definition and its relation to linear acoustic echo cancellation are illuminated.

Based on this approach, and assuming a converged linear echo canceller, a suppression approach for nonlinearly distorted acoustic echo signal components is introduced, the essential component of which is the nonlinear echo power spectral density estimation, which depends on one unknown real scalar parameter.

Subsequently, an identification algorithm for said parameter is presented, which results in a usable nonlinear echo suppressor still under the assumption of a converged linear echo canceller. Moreover, a generic comprehensive evaluation method for suppressor-type algorithms is advertised.

Finally, the challenge of concurrently adaptive linear echo canceller and nonlinear echo suppressor is analyzed and a solution for a full system is presented. After listening to several audio examples, the audience is invited to discuss the presentation’s contents.

Short biography

Ingo Schalk-Schupp studied systems engineering and technical cybernetics at Otto von Guericke University in Magdeburg, Germany. He graduated as a diploma engineer (Diplomingenieur) in 2012 with a diploma thesis titled “Speech Signal Enhancement in Automotive Environments” composed at Nuance Communications in Ulm, Germany. The thesis comprised two patent applications and was granted the best thesis award by the “Magdeburger Kybernetiker e.V.” Since 2012, he has been a PhD student with Nuance in Ulm and is supervised by Professor Andreas Wendemuth, Chair of Cognitive Systems at Otto von Guericke University Magdeburg. This presentation reflects the author’s findings from his PhD research.

Sonderkolloquium, Dr. Chidambaram Amalai, ETH Zürich / am 10.04.2017

10.04.2017 von 15:00 bis 16:30

Ludewig-Meyn-Str. 2. Raum Ü2/K (LMS2, R. Ü2/K)

Titel: Algorithmic Advances in Allocation and Scheduling

Abstract: We study the restricted case of Scheduling on Unrelated Parallel
Machines. In this problem, we are given a set of jobs J with processing
times p_j and each job may be scheduled only on some subset of machines
S_j. The goal is to find an assignment of jobs to machines to minimize
the time by which all jobs can be processed. In a seminal paper,
Lenstra, Shmoys, and Tardos designed an elegant 2-approximation for the
problem in 1987. The question of whether approximation algorithms with
better guarantees exist for this classic scheduling problem has since
remained a source of mystery.

In recent years, with the improvement of our understanding of
Configuration LPs, it now appears an attainable goal to design such an
algorithm. Our main contribution is to make progress towards this goal.
When the processing times of jobs are either 1 or epsilon < 1, we design
an approximation algorithm whose guarantee tends to 1+sqrt(3)/2 =
1.8660254, for the interesting cases when epsilon approaches 0. This
improves on the 2 - epsilon_0 guarantee recently obtained by
Chakrabarty, Khanna, and Li for some constant epsilon_0 > 0.

Kolloquiumsvortrag, M. Sc. Alexander Mäcker, Universität Paderborn / am 21.04.2017

21.04.2017 von 14:15 bis 15:45

Institut für Informatik, Ludewig-Meyn-Straße 2, 24118 Kiel, Raum: Übungsraum 2/K

Titel: Non-Clairvoyant Scheduling to Minimize Max Flow Time on a Machine with Setup Times

Abstract:

Consider a problem in which $n$ jobs that are classified into $k$ types arrive over time at their release times and are to be scheduled on a single machine so as to minimize the maximum flow time. The machine requires a setup taking $s$ time units whenever it switches from processing jobs of one type to jobs of a different type. We consider the problem as an online problem where each job is only known to the scheduler as soon as it arrives and where the processing time of a job only becomes known upon its completion (non-clairvoyance). We analyze a simple modification of the FIFO strategy and show the competitiveness to be $\Theta(\sqrt{n})$, which is optimal for greedy-like'' algorithms. We will then also consider a smoothed analysis of the competitiveness. The smoothed competitiveness turns out to only be $O(\varepsilon^{-2} \log^2 n)$ when processing times $p_j$ are independently perturbed by adding a random value uniformly drawn from $[-\varepsilon p_j, \varepsilon p_j]$, $0 < \varepsilon < 1$. The talk is based on joint work with Manuel Malatyali and Sören Riechers.

Kolloquiumsvortrag, Prof. Massimo Guarnieri, University Padua (Italien) / am 24.04.2017

24.04.2017 von 17:15 bis 18:45

Institute Ostufer, Geb. D, "Aquarium", Kaiserstr. 2, 24143 Kiel

Titel: Vanadium Redox Flow Batteries: Potentials and Challenges of an Emerging Storage Technology.

Kolloquiumsvortrag, Prof. Knut Graichen, Institut für Mess-, Regel- und Mikrotechnik, Universität Ulm / am 15.05.2017

15.05.2017 von 17:15 bis 18:45

Institute Ostufer, Geb. D, "Aquarium", Kaiserstr. 2, 24143 Kiel

Titel: Eingebettete nichtlineare Optimierung und MPC im (Sub-)Millisekundenbereich

Abstract: Nichtlineare Optimierungsverfahren und deren Anwendung im Zusammenhang mit der nichtlinearen modellprädiktiven Regelung (NMPC) sind in der Regelungstechnik insbesondere für dynamische Systeme mit mehreren Stellgrößen und zur Berücksichtigung von Systembeschränkungen von zunehmender Bedeutung. Eine Herausforderung bei der Umsetzung dieser Verfahren in der Praxis ist jedoch der hohe numerische Aufwand und die algorithmische Komplexität, insbesondere bei hochdynamischen Systemen mit Abtastzeiten im (Sub-)Millisekundenbereich. Die Problematik der Umsetzbarkeit wird durch die Tatsache weiter verschärft, dass Hardware-Lösungen in der industriellen Praxis häufig sehr limitierte Ressourcen besitzen. Als Beispiele seien Speicherprogrammierbare Steuerungen (SPS) oder Steuergeräte (Electronic Control Unit – ECU) genannt. Diesen Herausforderungen kann im Sinne eines eingebetteten Entwurfs durch die Verwendung von zugeschnittenen Optimierungsalgorithmen in Kombination mit einer echtzeitfähigen Auswertung begegnet werden.
Im Hinblick auf die oben genannten Herausforderungen präsentiert der Vortrag einen Ansatz zur nichtlinearen dynamischen Optimierung und modellprädiktiven Regelung, der eine echtzeitfähige Umsetzung selbst auf leistungsschwacher Hardware ermöglicht. Neben der methodischen und algorithmischen Vorstellung wird das Verfahren anhand von ausgewählten mechatronischen Beispielen erläutert.

Kolloquiumsvortrag, nn / am 19.05.2017

19.05.2017 von 14:45 bis 15:45

Institut für Informatik, Ludewig-Meyn-Straße 2, 24118 Kiel, Raum: Übungsraum 2/K

Kolloquiumsvortrag, Dr. Barth / am 22.05.2017

22.05.2017 von 17:15 bis 18:45

Institute Ostufer, Geb. D, "Aquarium", Kaiserstr. 2, 24143 Kiel

Titel: Hochfrequenz-Filterdesign unter Berücksichtigung hoher Transmissionsleistung

Abstract: Das Design von Hohlleiterfiltern mit Hilfe der Softwareprogramme HFSS, MATHCAD und AWR wird beschrieben. Das Design berücksichtigt die hohe Transmissionsleistung (hier 40KW gepulst), die bei Radargeräten auftritt und den Entwickler vor besondere Herausforderung stellt. Die Designmethode ist äußerst präzise, so dass die Anforderungen an das Filter schon im ersten Designschritt erreicht werden. Eine Abstimmung, die Feldkonzentrationen und damit Überschläge in den Filterresonatoren hervorrufen könnten, ist damit nicht nötig. Trotzdem wird eine Abstimmmethode bei hohen Feldstärken diskutiert.

Ein Filter, das für die Luftraumüberwachung eines Flughafens konstruiert wurde und dort auch seit Jahren in Betrieb ist, wird als Beispiel vorgestellt.

Kolloquiumsvortrag, nn / am 26.05.2017

26.05.2017 von 14:15 bis 15:45

Institut für Informatik, Ludewig-Meyn-Straße 2, 24118 Kiel, Raum: Übungsraum 2/K

Kolloquiumsvortrag, nn / am 29.05.2017

29.05.2017 von 17:15 bis 18:45

Institute Ostufer, Geb. D, "Aquarium", Kaiserstr. 2, 24143 Kiel

Kolloquiumsvortrag, Prof. Dr. Helmut Harbrecht, Uni Basel / am 02.06.2017

02.06.2017 von 14:15 bis 15:45

Institut für Informatik, Ludewig-Meyn-Straße 2, 24118 Kiel, Raum: Übungsraum 2/K

Titel: On multilevel quadrature for elliptic partial differential equations with random input

Abstract: This talk is dedicated to multilevel quadrature methods for the rapid solution of partial differential equations with a random input parameter. The key idea of such approaches is a sparse-grid approximation of the occurring product space between the stochastic and the spatial variable. We develop the mathematical theory and present error estimates for the computation of the solution’s moments with focus on the mean and the variance in case of second order elliptic boundary value problems with random diffusion. In particular, the present framework covers the multilevel Monte Carlo method and the multilevel quasi-Monte Carlo method as special cases. The theoretical findings are supplemented by numerical experiments.

Sonderkolloquium, Prof. Dr. Raj, Universität Agra, Indien / am 08.06.2017

08.06.2017 von 11:30 bis 13:00

Institut für Informatik, Christian-Albrechts-Platz 4, Raum: 910 (CAP 4), 24118 Kiel

Titel: Finite Element Modelling and Simulation of Selected Bulk Nano-material Processing Techniques

Abstract: A brief review of the material modeling such as visco-plastic, coupled thermo visco-plastic along with mathematical formulation is presented with application to forming of Bulk Nano-structured Materials using Severe Plastic Deformation(SPD). The material is assumed to be homogeneous, isotropic, incompressible. Elasticity is neglected and the material behavior chosen is given by the Norton-Hoff law. A power law is used to simulate the friction between the die and the work-piece. The flow problem with a given temperature distribution is solved simultaneously with the heat diffusion equation. Using the Penalty approach to enforce approximate incompressibility, the Variational principle is applied. This finds the velocity field solution of the problem by minimizing the functional. The coupling between the mechanical and the thermal problems is obtained through the dependency of the consistency of the material with temperature. Being a non-steady state process, SPD is analyzed by using small steady-like deformation steps. The usual finite element method is used to discretize the two problems at a given state Ω(t), and in the axi-symmetrical case the interpolation of the variables is performed in the (r2, z) space. The SPD processes that are modeled are Equal Channel Angular Pressing (ECAP), Twist Extrusion (TE), Accumulative Roll Bonding(ARB), Repetitive Corrugation and Straightening (RCS) and High Pressure Torsion (HPT). Finite Element Modeling and experimentation of these SPD processes is presented in this paper.

Kolloquiumsvortrag, PhD Tammy Riklin Raviv, Ben-Gurion-University of the Negev, Beer Sheva, Israel / am 09.06.2017

09.06.2017 von 13:30 bis 15:00

Institut für Informatik, Ludewig-Meyn-Straße 2, 24118 Kiel, Raum: Übungsraum 2/K

Titel:

Ensemble of Expert Deep Neural Networks for Spatio-Temporal Denoising of Contrast-Enhanced MRI Sequences

Abstract:

Abstract: Dynamic contrast-enhanced MRI (DCE-MRI) is an imaging protocol where MRI scans are acquired repetitively throughout the injection of a contrast agent. The analysis of dynamic scans is widely used for the detection and quantification of blood-brain barrier (BBB) permeability. Extraction of the pharmacokinetic (PK) parameters from the DCE-MRI concentration curves allows quantitative assessment of the integrity of the BBB functionality. However, curve fitting required for the analysis of DCE-MRI data is error-prone as the dynamic scans are subject to non-white, spatially dependent and anisotropic noise.

We present a novel spatio-temporal framework based on Deep Convolution Neural Networks (DCNNs) to address the DCE-MRI denoising challenges. This is accomplished by an ensemble of expert DCNNs constructed as deep autoencoders, where each is trained on a specific subset of the input space to accommodate different noise characteristics and curve prototypes. The most likely reconstructed curves are then chosen using a classifier DCNN followed by a quadratic programming optimization. As clean signals (ground-truth)

for training are not available, a fully automatic model for generating realistic training sets with complex nonlinear dynamics is introduced. The proposed approach has been applied to full and even temporally down-sampled DCE-MRI datasets acquired by MRI machines in different locations and of different manufacturers and is shown to favorably compare to state-of-the-art denoising methods.

Kolloquiumsvortrag, Prof. Herbert Jäger, Jacobs University Bremen / am 12.06.2017

12.06.2017 von 17:15 bis 18:45

Institut für Elektrotechnik und Informationstechnik, Geb. D, "Aquarium", Kaiserstr. 2, 24143 Kiel

Thema: An introduction to Reservoir Computing

Abstract: Recurrent neural networks (RNNs) are general approximators for nonlinear dynamical systems and have recently become widely used in the "deep learning" field of machine learning, especially for speech and language processing tasks. For instance, Google's speech recognition and language translation services are based on RNNs.

However, the deep learning set-ups for RNN training are computationally very expensive, require very large volumes of training data, and need high-precision numerical processing. For such reasons, deep-learning variants of RNNs are problematic in fields where training data are scarce, where fast and cheap algorithms are desired, or where noisy or low-precision hardware is to be used. This is often the case in domains of nonlinear signal processing, control, brain-machine interfacing, or biomedical signal processing.

Reservoir Computing (RC) is an alternative machine learning approach for RNNs which is in many aspects complementary to the ways of deep learning. In RC, a large, random, possibly low-precision and noisy RNN is used as a nonlinear excitable medium - called the "reservoir" - which is driven by an input signal. The reservoir itself is not adapted or trained. Instead, only a "readout" mechanism is trained, which assembles the desired output signal from the large variety of random, excited signals within the reservoir. This readout training is cheap - typically just a linear regression. RC has become a popular approach in research that aims at useful computations on the basis on unconventional hardware (non-digital, noisy, low-precision).

The talk gives an introduction to the basic principles and variants of RC. Numerous examples will be presented according to wishes from the audience.

Sonderkolloquium, Prof. Remus Teodorescu, Universität Aalborg / 12.06.2017

12.06.2017 von 15:15 bis 16:45

Institute Ostufer, Geb. D, "Aquarium", Kaiserstr. 2, 24143 Kiel

Titel: Modular Multilevel Converter Research at Aalborg University

Abstract:This presentation is aiming on summarizing the main features of the recently published book: Design, Control, and Application of Modular Multilevel Converters for HVDC Transmission Systems, ISBN: 978-1-118-85156-2, Wiley 2016http://eu.wiley.com/WileyCDA/WileyTitle/productCd-1118851560.html#. Special focus will be on Control under unbalanced conditions of MMC.

Kolloquiumsvortrag, nn / am 16.06.2017

16.06.2017 von 14:15 bis 15:45

Institut für Informatik, Ludewig-Meyn-Straße 2, 24118 Kiel, Raum: Übungsraum 2/K

Kolloquiumsvortrag, Prof. Dr.-Ing. Mücklich, Functional Materials, Dept. Mat. Science & Engineering, Saarland University / am 19.06.2017

19.06.2017 von 17:15 bis 18:45

Institute Ostufer, Geb. D, "Aquarium", Kaiserstr. 2, 24143 Kiel

Titel: Optimized functionality by 3D microstructure research and efficient surface patterning

Abstract: In the introduction a short overview will be given concerning the three units of the institute, such as the chair for functions materials which is dedicated to fundamental research and teaching, the European School of Materials, which is focused on the international study programs on all academic levels and the Material Engineering Center Saarland, which is promoted by applied research and transfer activities. Then some of the research activities will be discussed based on three main questions:

-          How does 3D microstructure research on the micro, nano and atomic scale help to understand the quantitative relations between microstructure formation and properties

-          How does the initial 3D morphology control the processing and formation of microstructures and

-          How can Direct Laser Interference Patterning be exploited to design optimized surface functionalities

Sonderkolloquium Prof. Tamulevicius, Technische Universität Kaunas / am 20.06.2017

20.06.2017 von 17:15 bis 18:45

Technische Fakultät, Institut für Materialwissenschaft, Kaiserstr. 2, Kiel, Raum: "Aquarium", Geb. D

Titel: Nanoparticles, nanocomposites - from optics to medicine

Abstract: Principles of deposition and applications of diamond-like carbon thin films, diamond-like carbon-based nanocomposites including metallic nanoparticles as well as capillary assisted deposition of nanoparticles will be presented, thereby concentrating on the optical, electrical  properties and the use of the assembly of particles as building blocks for optical sensors and antimicrobial surfaces. Features of localized surface plasmon effects, surface enhanced Raman scattering, detection of ultrafast energy transfer processes will be discussed.

Kolloquiumsvortrag, nn / am 23.06.2017

23.06.2017 von 14:15 bis 15:45

Institut für Informatik, Ludewig-Meyn-Straße 2, 24118 Kiel, Raum: Übungsraum 2/K

Sonderkolloquium, Prof. R.D. Lorenz, University of Wisconsin-Madison / 23.06.2017

23.06.2017 von 13:00 bis 15:00

Institute Ostufer, Geb. D, "Aquarium", Kaiserstr. 2, 24143 Kiel

Titel: Disruptive Paradigm Changes for Electrical Machines and Electrical Drives

Abstract: Electric machine design paradigms have been dramatically changed by the need to meet the demands for minimizing losses and smooth torque control during driving cycles with widely vary loads and speeds.  Simultaneously, new drive control paradigms systematically out-perform industry standard field oriented control (FOC) and simultaneously solve several classical problems with FOC. In addition, the internet of things is opening expansive opportunities for motor drives.  This presentation with focus on these disruptive changes in the paradigms for electric machines and electrical drives and explore the opportunities for innovation that these technologies provide.

Affiliation:

Chaired Professor and Co-Director of WEMPEC at the University of Wisconsin-Madison, in Madison, Wisconsin, USA

Short Bio:

Prof. Robert D. (Bob) Lorenz is a Life Fellow of IEEE, Past President of IEEE IAS, and Past Member of the IEEE Board of Directors. He is a Chaired Professor at the University of Wisconsin-Madison and Co-Director of the Wisconsin Electric Machines and Power Electronics Consortium (WEMPEC) which just celebrated its 36th Anniversary with over 85 international sponsor firms.    He has pioneered core technologies for physics-based control design, self-sensing, flux observers, current regulators, deadbeat-direct torque and flux control, variable flux and variable magnetization state PM machines and power semiconductor temperature and strain control and has won 33 prize paper awards from the IEEE.

Kolloquiumsvortrag, Antrittsvorlesung Prof. Kapels / am 26.06.2017

26.06.2017 von 17:15 bis 18:45

Institute Ostufer, Geb. D, "Aquarium", Kaiserstr. 2, 24143 Kiel

Titel: Von der Siliziumtechnologie zu Wide-Bandgap Leistungshalbleitern

Abstract: Kontinuierliche Innovationen im Bereich der siliziumbasierten Leistungshalbleiterbauelemente ermöglichten grundlegende Fortschritte für eine effiziente dezentrale Energieversorgung, die Weiterentwicklung der Elektromobilität und energieeffiziente Stromrichter für Industrie- und Consumer-Anwendungen. Aktuelle Trends in Leistungsdichte und Gewicht werden jedoch zunehmend nicht mehr mit Halbleiterbauelementen auf Siliziumbasis erfüllt werden können. Leistungstransistoren auf Basis von SiC und GaN ermöglichen hier Zukunftspotentiale zu heben. Die Vorlesung zeigt die wesentlichen Bauelementekonzepte, aktuelle Herausforderungen und Lösungen auf.

Sommerfest am 30.06.2017

30.06.2017 von 15:00 bis 15:00

Technische Fakultät, Kaiserstr. 2, 24143 Kiel

Kolloquiumsvortrag, Carsten Grashoff / am 03.07.2017

03.07.2017 von 17:15 bis 18:45

Institute Ostufer, Geb. D, "Aquarium", Kaiserstr. 2, 24143 Kiel

Titel: Piconewton-sensitive biosensors to investigate molecular forces in cells

Abstract: The ability of cells to adhere and simultaneously sense differences in tissue stiffness is crucial for organ development and function. Yet, the molecular mechanisms by which cells sense extracellular matrix rigidity have remained unknown because suitable techniques to measure mechanical forces across intracellular proteins in living cells were missing.

We therefore develop novel, single-molecule‒calibrated tension sensor modules that allow the analysis of a physiologically highly relevant force regime in cells. Our new probes are sensitive to forces of 3–5 piconewton (pN), 6–8 pN and 9–11 pN, respectively; they are characterized by fast folding/unfolding transitions, reversibility and a sharp force-response threshold (1, 2, 3). By applying these new probes to the cell adhesion proteins talin-1 and talin-2, we demonstrate that these central integrin activators establish intracellular, mechanical linkages that bear mechanical forces of about 7–10 pN upon cell adhesion and are regulated by f-actin and vinculin association. We find that the integrin–talin–actin linkage is indispensable for extracellular rigidity sensing and, surprisingly, talin isoform-specific (1). Furthermore, multiplexing distinct tension sensor constructs by dual-color FLIM reveals an unexpected intramolecular tension gradient across talin that is modulated by intracellular signals and extracellular rigidity (3).

References:

1. Austen KA et al., and Grashoff C. 2015. Nat Cell Biol.

2. Freikamp A, Cost AL, and Grashoff C. 2016. Trends Cell Biol.

3. Ringer et al., and Grashoff C. 2017. under review

Sonderkolloquium, Jun.-Prof. Dr. Mattias Heinrich, Uni Lübeck / am 07.07.2017

07.07.2017 von 14:15 bis 15:45

Institut für Informatik,Christian-Albrechts-Platz 4, R.715, 24114 Kiel

Titel: Learning Sparse Binary Features for Medical Image Segmentation of the Abdomen

Abstract: In this talk, we explore the capabilities of sparse binary features for medical image segmentation. Due to insufficient contrast and anatomical shape variations local image patches rarely provide sufficient information for accurate segmentation of abdominal structures. Based on our two recent MICCAI papers, we propose to use long-range binary features to robustly capture the image context. Two different classification strategies are subsequently developed.

First, a very fast approximate nearest neighbour search based on vantage point forests and Hamming distances between feature strings is presented. The classifier can be learned and applied to new data in few seconds. The approach reaches state-of-the-art performance for larger organs on the VISCERAL3 benchmark.

Second, we develop a deep neural network architecture that combines a local CNN path with a new contextual path that encodes the sparse binary features. Following the ideas from Network-in-Network, 1x1 convolutions are employed to learn the best combination of different binary offset locations. We demonstrate experimentally that this restricted feature extraction in the first layer enables to regularise the network with a huge receptive field and leads to short training times of less than 10 minutes. Using only 1 million trainable parameters, the model achieves a accuracy of 64.5% Dice, which is comparable to the best performing, much more complex deep CNN approach for pancreas segmentation.

Finally, the potential use of learned binary features for other tasks in medical image analysis, such as image registration and disease classification will be discussed.

Kolloquiumsvortrag, Prof. Grandoni, IDSIA USI-SUPSI in Lugano / am 07.07.2017

07.07.2017 von 14:15 bis 15:45

Institut für Informatik, Ludewig-Meyn-Str. 2, Raum Ü2/K (LMS2, R. Ü2/K), 24114 Kiel

Titel: Approximating Geometric Knapsack via L-packings

Abstract: Joint work with: Waldo Galvez, Sandy Heydrich, Salvatore Ingala,
Arindam Khan, Andreas Wiese

In the 2-dimensional geometric knapsack problem (2DK) we are given a
set of n axis-aligned rectangular items, each one with an associated
profit, and an axis-aligned square knapsack. The goal is to find a
(non-overlapping) packing of a maximum profit subset of items inside
the knapsack (without rotating items). The best-known polynomial-time
approximation factor for this problem (even just in the cardinality
case) is 2 + ε [Jansen and Zhang, SODA 2004]. In this work we break
the 2 approximation barrier, achieving a polynomial-time 17/9 + ε <
1.89 approximation, which improves to 558/325+ ε < 1.72 in the
cardinality case.

Essentially all prior work on 2DK approximation packs items inside a
constant number of rectangular containers, where items inside each
container are packed using a simple greedy strategy. We deviate for
the first time from this setting: we show that there exists a large
profit solution where items are packed inside a constant number of
containers plus one L-shaped region at the boundary of the knapsack
which contains items that are high and narrow and items that are wide
and thin. The items of these two types possibly interact in a complex
manner at the corner of the L.

The above structural result is not enough however: the best-known
approximation ratio for the sub-problem in the L-shaped region is 2 +
ε (obtained via a trivial reduction to 1-dimensional knapsack by
considering tall or wide items only). Indeed this is one of the
simplest special settings of the problem for which this is the best
known approximation factor. As a second major, and the main
algorithmic contribution of this work, we present a PTAS for this
case. We believe that this will turn out to be useful in future work
in geometric packing problems.

We also consider the variant of the problem with rotations (2DKR),
where items can be rotated by 90 degrees. Also in this case the
best-known polynomial-time approximation factor (even for the
cardinality case) is 2 + ε [Jansen and Zhang, SODA 2004]. Exploiting
part of the machinery developed for 2DK plus a few additional ideas,
we obtain a polynomial-time 3/2 + ε-approximation for 2DKR, which
improves to 4/3 + ε in the cardinality case.

Kolloquiumsvortrag, nn / am 10.07.2017

10.07.2017 von 17:15 bis 18:45

Institute Ostufer, Geb. D, "Aquarium", Kaiserstr. 2, 24143 Kiel

Sonderkolloquium, Tomislav Dragicevic, University Aalborg / 12.07.2017

12.07.2017 von 15:00 bis 16:45

Institute Ostufer, Geb. D, "Aquarium", Kaiserstr. 2, 24143 Kiel

Titel: Coordination Strategies for DC Microgrids

Abstract: DC distribution systems have higher efficiency, better current carrying capacity and faster response when compared to conventional AC systems. They also provide more natural interface with many types of RES and ESSs and better compliance with consumer electronics. Furthermore, when components are coupled around a DC bus, there are no issues with reactive power flow, power quality and frequency regulation, resulting in a notably less complex control system when compared to the AC coupled systems. All these facts lead to more and more applications of DC systems in modern power systems, including data/telecom centers, maritime industry, high voltage transmission systems, electric vehicle charging infrastructure, and DC microgrids. Still, design and operation of general DC systems imposes a number of specific challenges. The aim of this lecture is in particular to present functionalities of different coordinated control strategies for DC microgrids. In this context, centralized, decentralized and distributed controls are assessed. Decentralized control can be regarded as an extension of local control since it is also based exclusively on local measurements. In contrast, centralized and distributed control strategies rely on digital communication technologies. A number of approaches of using these three coordinated control strategies to achieve various control objectives are reviewed in the paper. Moreover, properties of DC MG dynamics and stability are discussed.

Kolloquiumsvortrag, nn / am 14.07.2017

14.07.2017 von 14:15 bis 15:45

Institut für Informatik, Ludewig-Meyn-Straße 2, 24118 Kiel, Raum: Übungsraum 2/K

Kolloquiumsvortrag, nn / am 17.07.2017

17.07.2017 von 17:15 bis 18:45

Institute Ostufer, Geb. D, "Aquarium", Kaiserstr. 2, 24143 Kiel

Sonderkolloquium, Prof. Dr. María López Fernández, Universität Zürich / am 18.07.2017

18.07.2017 von 11:00 bis 13:00

Ludewig-Meyn-Str. 2. Raum Ü2/K (LMS2, R. Ü2/K), 24118 Kiel

established method for the time discretization of retarded potentials
associated to wave equations. It has been very much developed in the
last decade, both from the theoretical and the algorithmic point of
view. However, despite its nice properties, the Convolution Quadrature
is strictly restricted to the use of fixed time steps. In this talk I
will present the "generalized Convolution Quadrature", a new family of
methods designed to overcome the strong restriction to uniform
temporal grids. I will show stability and convergence estimates and
numerical results illustrating the good behaviour of the new method. I
will also outline the current limitations in the implementation of the
generalized Convolution Quadrature and future possibilities of
development.

Kolloquiumsvortrag, Prof. Dr. Stefan Sauter, Uni Zürich / am 21.07.2017

21.07.2017 von 14:15 bis 15:45

Institut für Informatik, Ludewig-Meyn-Straße 2, 24118 Kiel, Raum: Übungsraum 2/K

Titel: A Family of Crouzeix-Raviart Non-Conforming Finite Elements in Two- and Three Spatial Dimensions

Abstract: In this talk we will present a family of non-conforming "Crouzeix-Raviart" type finite elements in two and three dimensions. They consist of local polynomials of maximal degree p on simplicial finite element meshes while certain jump conditions are imposed across adjacent simplices.

We will prove optimal a priori estimates for these finite elements. The characterization of this space via jump conditions goes back to the seminal paper of Crouzeix and Raviart in 1973. However, the definition is implicit and the derivation of an explicit representation of the local basis functions for general p in 3D was an open problem.

We present explicit representations for these functions by developing some theoretical tools for fully symmetric and reflection symmetric orthogonal polynomials on triangles and their representation.

Finally we will analyze the linear independence of these sets of functions and discuss the question whether they span the whole non-conforming space. This talk comprises joint work with P. Ciarlet Jr., ENSTA, Paris and Charles F. Dunkl, Virginia Tech.