Wissenschaftliches Kolloquium

Prof. Dr. Karl-Heinz Zimmermann gives a presentation on

Computing using Molecules

on November 8th, 2017 at 15.30 PM in room N.0010.

 

Abstract:

DNA computing is a form of data processing where the genetic molecule DNA serves as data carrier such that the DNA-encoded data can be processed with bio-technological operations. Using DNA computers, activities within a cell can be controlled in-vivo, as well as DNA can be processed outside cells in-vitro and preferably autonomously. Furthermore, models for DNA computing exist that are interesting for data processing in-silico. This talk briefly summarizes different approaches to DNA computing.

Short Bio of the speaker:

  • Studies of computer science and mathematics, University of Erlangen, Germany
  • PhD degree in theoretical computer science, University of Erlangen, Germany
  • Habilitation in mathematics, University of Bayreuth, Germany
  • Fulbright scholarship, Princeton University, USA
  • Heisenberg scholarship, Karlsruhe Institute of Technology, Germany
  • Professorship in computer science, Hamburg University of Technology
  • Visiting professor, Cornell and Princeton Universities, USA

All interested people are kindly invited.


Prof. Dr. Chris Brzuska gives a presentation on

Computational hardness and its bright side

on October 11th, 2017 at 15.30 PM in room N.0010.

 

Abstract:

SAT-solvers solve large practical instances every day. It is thus tempting to think of P vs. NP as a question of purely theoretical nature that does not generate any hardness in practice. Luckily, the P vs. NP problem also seems to generate practical hardness in the form of secure cryptography. But how to we find such useful, hard instances? Indeed - what makes a problem hard? What makes a problem easy? And what makes a problem useful for cryptography?

Research on random constraint satisfaction (CSP) problems tells us that (non-surprisingly) hardness of random instances depends on the ratio between the number of constraints and the number of variables. Interestingly, the hardness seems to appear and disappear suddenly, at particular constraint-to-variable ratios. The most promising avenue to understanding the experimental hardness results is via proving that at the experimental thresholds, fundamental properties of the solution space change. Random CSPs are a traditional area of computer science, ideas for the threshold phenomena emerge from physics, and rigorous proofs (if existent) of those ideas often involve mathematicians. The hope of such an interdisciplinary field is that it might uncover some foundational explanations of computational hardness and thus provide a solid foundation for cryptography. In the talk, we will discuss some of those hopes.

Short Bio of the speaker:

  • 2005 - 2010 Studies of mathematics at the University of Duisburg-Essen, the university of Bordeaux and TU Darmstadt
  • 2012 PhD in computer science at TU Darmstadt, including 6 months at the Institute of Advanced Studies in Princeton, USA
  • 2012 - 2014 Postdoctoral researcher at Tel-Aviv University, Isreael
  • 2014 - 2015 Postdoctoral researcher at Microsoft Research Cambridge, UK
  • 2015 Junior professor for IT Security Analysis, Hamburg University of Technology

All interested people are kindly invited.


Prof. Dr. Wolfgang Krautschneider gives a presentation on

Electronic Implants for Medical Applications

on September 13th, 2017 at 15.30 PM in room N.0010.

 

Abstract:

Integrated circuits with advanced CMOS nanoelectronic technology offer great potential for medial applications, especially for implants. For this purpose, the integrated circuits have to be tailored to the special requirements of in-vivo usage including small area and low power consumption. As commercial chips for these special applications are not available, chips that fit these needs, i.e., Application Specific Integrated Circuits (ASICs) have to be designed. This is done by computer simulations in the institute. After applying several computer based check routines, the ASICs are fabricated in silicon foundries and then sent back to the institute for hardware testing and characterization. When the ASIC fulfills the specifications the implant can be manufactured.

The development of ASICs and the build-up of medical implants will be illustrated based on projects in the field of myoelectric prosthesis, aneurysm of the aorta and cancer therapy.

Short Bio of the speaker:

  • 1977 Diploma in Electrical Engineering at TU Berlin
  • 1982 PhD at TU Berlin
  • 1984 Postdoctoral Fellowship at IBM Central Research Lab, Yorktown Heights, NY, USA
  • 1986 Siemens AG, Munich, Megabit Project
  • 1990 IBM-Siemens Cooperation, Burlington, VT, USA
  • 1993 Siemens AG, Munich, Gigabit memory chip development
  • 1997 Habilitation, TU Berlin
  • 1999 Professor at Institute of Nano and Medical Electronics, Hamburg University of Technology
  • 2008 Member of Berlin-Brandenburg Academy of Science

All interested people are kindly invited.


Dr. Patricio Farrell gives a presentation on

Solving Partial Differential Equations Numerically: Meshfree and Finite Volume Methods

on July 12th, 2017 at 15.30 PM in room N.0010.

 

Abstract:

How do you solve a partial differential equation on a computer? In this talk, we give an accessible introduction to two helpful numerical discretization tools: Radial basis functions (RBFs) and finite volume methods (FVMs). The former allow to solve differential equations without the cumbersome generation of a grid. The latter are particularly useful in the context of semiconductor device simulation by preserving (discrete) properties which the continuous system satisfies.

Short Bio of the speaker:

  • 2005 - 2010 Studies at University of Hamburg
  • 2010 Diploma degree, University of Hamburg
  • Since 2007 Freelance journalist
  • 2010 - 2014 PhD studies at University of Oxford
  • 2014 - 2017 Postdoctoral researcher, Weierstrass Institute Berlin
  • Since 2017 Visiting professor at Hamburg University of Technology

All interested people are kindly invited.


Prof. Dr.-Ing. Gerhard Bauch gives a presentation on

Information Theory for Computationally Efficient Signal Processing Implementation

on June 14th, 2017 at 15.30 PM in room N.0010.

 

Abstract:

The core of information theory is the derivation of fundamental limits of information compression and transmission. Information theory focuses on asymptotic results in the sense of infinitely long blocks of information bits which have to be stored or transmitted. Aspects which are essential in real world systems such as signal processing complexity, memory   requirements or delay are usually not taken into account in information theoretic considerations. Nevertheless, information theory has become a fundamental tool for the design of sophisticated modern communications systems. A prominent example is the application of multiple antenna technology (MIMO) for achieving high spectral efficiency in the LTE system: The gains of MIMO are all but intuitive but are motivated by results from information theory.

In our current research, we go even a step further and use information theory for the design of computationally efficient detector software or even hardware implementations – even though the spirit of information theory actually totally neglects implementation aspects. The exciting observation is that this approach has the potential to even outperform state of the art implementations while providing reduced complexity and higher detection speed – a combination which is rarely seen in engineering solutions.

The talk tries to illuminate the basic idea using the example of forward error control decoders and sketches some further application areas we are working on.

Short Bio of the speaker:

  • 1995 Dipl.-Ing. degree, Electrical Engineering, Munich University of Technology
  • 1996 German Aerospace Center (DLR), Oberpfaffenhofen, Germany
  • 1996 - 2001 Research Assistant, Munich University of Technology
  • 1998 and 1999 Visiting Researcher AT&T Labs Research, Florham Park, NJ/USA
  • 2001 Dr.-Ing. degree, Munich University of Technology
  • 2001 Dipl.-Volksw. degree, Economics, FernUniversität Hagen
  • 2002 - 2008 Head Advanced Radio Transmission Group, DOCOMO Euro-Labs, Munich, Germany
  • 2007 Research Fellow of DOCOMO Euro-Labs, Munich, Germany
  • 2009 - 2012 Full professor at Universität der Bundeswehr, Munich, Germany
  • 2012 - now Head of Institute of Communications, Hamburg University of Technology

All interested people are kindly invited.


Prof. Dr. Alexander Schlaefer gives a presentation on

Machine Learning for High Resolution Image Guidance

on May 10th, 2017 at 15.30 PM in room N.0010.

 

Abstract:

Precise localization of target structures inside a patient is a key requirement for many clinical applications, including minimally invasive surgery. Typical trade-offs include spatial resolution, temporal resolution, and field of view. We consider optical coherence tomography (OCT), which combines an imaging depth of 1-2 mm in soft tissue with high spatial and temporal resolution at approximately 10 μm and 1.5 MHz, respectively. Challenges include the need for online processing at high data rates and speckle noise complicating image processing. The talk will illustrate how OCT can be used for real time tracking of motion and deformation. Particularly, we will show how machine learning based approaches can be used to identify pose, interaction force, and tissue type from OCT image data.

Short Bio of the speaker:

  • 2001 Dipl.-Inf. degree, Artificial Intelligence Workgroup, Humboldt Universität zu Berlin
  • 2003 - 2007 Research Assistant, Institute for Robotics and Cognitive Systems, Universität zu Lübeck
  • 2007 Dr.-Ing. degree, Universität zu Lübeck
  • 2007 - 2008 Postdoctoral Research Scholar, Department of Radiation Oncology, Stanford University, CA/USA
  • 2008 - 2013 Assistant Professor, Medical Robotics, Universität zu Lübeck
  • 2013 - now Professor, Institute of Medical Technology, Hamburg University of Technology

All interested people are kindly invited. (Foto: TUTECH/Marco Grundt)


Prof. Dr. Arne Jacob gives a presentation on

Packaging Approaches for Broadband Communication Systems

on April 12th, 2017 at 15.30 PM in room N.0010.

 

Abstract:

One remedy to satisfy the ever increasing need for bandwidth is to move communication systems to higher operating frequencies. Besides the challenges imposed by the decreasing wavelength and the associated miniaturization, various other design constraints may apply depending on the application. This has a strong impact on component and subsystem packaging. In this presentation, three examples are considered which illustrate possible approaches. The first one deals with a 2.5D circuit integration in LTCC (Low-Temperature Co-fired Ceramics) for satellite applications and with the on-orbit verification of the concept. The second example reports a package design for solid-state power amplifiers to be used in active multi-beam satellite antennas. At last, a low-cost packaging approach for 100 Gbps communication systems is sketched.

Short Bio of the speaker:

  • 1979 Dipl.-Ing. degree in Electrical Engineering, High-Frequency Technology, TU Braunschweig
  • 1986 Dr.-Ing. degree, TU Braunschweig
  • 1986 - 1988 "Fellow", RF Group, Super Proton Synchrotron, CERN, Geneva, Switzerland
  • 1988 - 1990 Staff Scientist, Accelerator and Fusion Research Division, Lawrence Berkeley Laboratory, Berkeley, CA/USA
  • 1990 - 2004 Professor of Microwave Engineering, TU Braunschweig
  • 2004 - now Professor of High-Frequency Technology, Hamburg University of Technology

All interested people are kindly invited.


Prof. Dr. Volker Turau gives a presentation on

Data Stream Processing in Wireless Networks

on March 8th, 2017 at 15.30 PM in room N.0010.

 

Abstract:

A data stream is a sequence of digitally encoded signals that represent continuously recorded information from a particular location. Embedded into the physical world, wireless sensor networks can aggregate such data streams from different locations into a single stream. Monitoring and analyzing large and rapidly changing streams of data that arrive online in an environment with with bounded resources (bandwidth, memory, etc.) is a challenging task. The unbounded length of a data stream makes it impossible to store the entire contents of the stream, yet many applications demand to retain some ability to execute queries referencing past data. The talk considers the problem of maintaining statistics over streams with regard to all or the last n data elements seen so far.

Short Bio of the speaker:

  • Academic studies at University of Mainz
  • Diploma degree in mathematics in 1984
  • PhD degree in mathematics in 1987 (subject: algebraic groups)
  • Postdoc positions in Manchester, Karlsruhe and Berkeley
  • Member of TUHH since September 2002

All interested people are kindly invited.