HGS MathComp Curriculum & Events

The compact course intents to give a broad overview of the financial industry, its instruments and markets with a quantitative focus. It addresses the valuation of standard financial instruments - e.g. options – and portfolio theory in more detail. After the course you should be able to decide whether to go deeper into a specific area of quantitative finance. Basics in differential-equations, numeric calculus and stochastic processes are helpful but not a must. Bring your laptop with you.

Please register here

Continuous and discrete signals, sampling theorem, signal representation
Fourier transform
Random variables and fields, probability density functions, error propagation
Homogeneous and inhomogeneous point operations
Neighborhood operations, linear and nonlinear filters, linear system theory
Geometric transformations and interpolation
Multi-grid signal presentation and processing
Averaging, edge and line detection, local structure analysis, local phase and wave numbers
Motion analysis in image sequences
Segmentation
Regression, globally optimal signal analysis, variation approaches, steerable and nonlinear filtering, inverse filtering
Morphology and shape analysis, moments, Fourier descriptors
Bayesian image restoration
Object detection and recognition

_____________________________

Learn how to analyze signals from time series, images, and any kind of multidimensional signals and to apply it to problems in natural sciences, life sciences and technology.
________________________

(MWInf6)

In this talk, we address several aspects of model and cost reduction techniques for the numerical simulation of PDEs. Firstly, we consider tearing and interconnection strategies in combination with interface models and reduced basis methods. These techniques are of special interest in the case of vibro-acoustic simulation and can largely reduce the number of degrees of freedom. Numerical results in case of orthotropic material laws are given for a violinbridge as well as a timber building.
Secondly, we discuss surrogate models which aim at a complexity reduction. Here we introduce locally mass conservative scheme in a local postprocessing step as well as adaptive control techniques for resilient multigrid solvers. In all cases, we provide the numerical analysis, algortihmic aspects and illustrative large scale results.

Human motion measurement and analysis is a challenging problem, due to issues such as sensor and measurement system limitations, high dimensionality, and spatial and temporal variability. Accurate and timely motion measurement and analysis enables many applications, including imitation learning for robotics, new input and interaction mechanisms for interactive environments, and automated rehabilitation monitoring and assessment. In this talk we will describe recent work in the Adaptive Systems Laboratory at the University of Waterloo developing techniques for automated human motion measurement and analysis. We will overview techniques for motion measurement, segmentation, individualized model learning and analysis, with a focus on two application areas: rehabilitation and interactive environments.

Biography:
Dana Kuli? received the combined B.A.Sc. and M.Eng. degrees in electromechanical engineering, and the Ph.D. degree in mechanical engineering from the University of British Columbia, Canada, in 1998 and 2005, respectively. From 2006 to 2009, she was a JSPS Postdoctoral Fellow and a Project Assistant Professor at the Nakamura Laboratory at the University of Tokyo. She is currently an Associate Professor at the Electrical and Computer Engineering Department at the University of Waterloo, Canada. She is a founding co-chair of the IEEE RAS Technical Committee on Human Movement Understanding and an Associate Editor with the IEEE Transactions on Robotics. In 2014, she was awarded Ontario’s Early Researcher award for her work on rehabilitation and human-robot interaction. Her research interests include human motion analysis, robot learning, humanoid robots, and human-machine interaction.

We consider the question of when delay systems, which are intrinsically infinite dimensional, can be represented by finite dimensional systems. Specifically, we give conditions for when all the information about the solutions of the delay system can be obtained from the solutions of a finite system of ordinary differential equations. For linear autonomous systems and linear systems with time-dependent input we give necessary and sufficient conditions and in the nonlinear case we give sufficient conditions. The ideas and results are illustrated by models for infectious diseases and physiologically structured populations.

Significant digitization efforts have resulted in large music collections, which comprise music-related documents of various types and formats including text, symbolic data, audio, image, and video. In the field of music information retrieval (MIR) great efforts are directed towards the development of technologies that allow users to access and explore music in all its different facets. For example, during playback of some CD recording, a digital music player may present the corresponding musical score while highlighting the current playback position within the score. On demand, additional information about melodic and harmonic progression or rhythm and tempo is automatically presented to the listener. A suitable user interface displays the musical structure of the current piece of music and allows the user to directly jump to any key part within the recording without tedious fast-forwarding and rewinding. In this talk, I discuss a number of current research problems in the field of music information retrieval and indicate possible solutions. One fundamental problem is to decompose a given music signal into semantically meaningful components. To guide the decomposition, one may exploit additional information, either in the form of specific acoustic properties of the components or in the form of additional score information. As an example, I show how to compute a notewise decomposition of a music signal by applying a score-informed variant of non-negative matrix factorization (NMF). Finally, I discuss various audio editing and manipulating applications to highlight the potential of these decomposition techniques.

Biography:
Meinard Müller studied mathematics (Diplom) and computer science (Ph.D.) at the University of Bonn, Germany. In 2002/2003, he conducted postdoctoral research in combinatorics at the Mathematical Department of Keio University, Japan. In 2007, he finished his Habilitation at Bonn University in the field of multimedia retrieval. From 2007 to 2012, he was a member of the Saarland University and the Max-Planck Institut für Informatik leading the research group "Multimedia Information Retrieval and Music Processing" within the Cluster of Excellence on "Multimodal Computing and Interaction". Since September 2012, Meinard Müller holds a professorship for Semantic Audio Processing at the International Audio Laboratories Erlangen, which is a joint institution of the Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU) and the Fraunhofer-Institut für Integrierte Schaltungen IIS. His recent research interests include music processing, music information retrieval, audio signal processing, and motion processing. Meinard Müller has been a member of the IEEE Audio and Acoustic Signal Processing Technical Committee from 2010 to 2015 and is a member of the Board of Directors of the International Society for Music Information Retrieval (ISMIR) since 2009. He has co-authored more than 100 peer-reviewed scientific papers, wrote a monograph titled "Information Retrieval for Music and Motion" (Springer, 2007) as well as a textbook titled "Fundamentals of Music Processing" (Springer, 2015, www.music-processing.de).

Ziel des Workshops ist es, die Hürden des Bewerbungsprozesses leichter zu nehmen und
vor allem im persönlichen Gespräch / Auswahlverfahren einen bleibenden
– positiven - Eindruck zu hinterlassen.

Bewerben ist wie flirten: Wer zu langweilig und 08/15 ist oder den falschen Ton trifft, wird abgewiesen. Daher gilt es in den richtigen Momenten zu punkten und sympathisch die Herausforderungen zu meistern. Gerade unter Druck gilt es souverän zu bleiben und Akzente zu setzen.

Das bedeutet beim Bewerben wie beim Flirten:
mehr Auswahl und einen höheren Ertrag.

Inhalt:

- Die typischen Stressfragen und die typischen Fehler des Bewerbers.
- Das Vorstellungsgespräch aus beiden Perspektiven beleuchtet.
- Worauf achtet ein Personaler?
- Wie viel Gehalt kann ich verlangen?

Nach dem Seminar werden die Teilnehmer:

- Im Vorstellungsgespräch punkten
- Aus der Masse hervorstechen.
- Im Stressinterview unter Druck cool bleiben.
- Gekonnt die Gehaltsfrage klären.
- Besser flirten

Bitte hier anmelden

The students
- can describe basic measures and characteristics of complex networks
- can implement and apply basic network analysis algorithms
- can describe different network models and can describe, compute, and analyze characteristic parameters of these models
- know how to compute different complex network measures and how to interpret these measures
- know different generative models for constructing complex networks, especially scale-free networks
- know the fundamental methods for the detection of communities in networks and the analysis of their evolution over time
- are familiar with basic concepts of network robustness
- understand the spread of phenomena in complex networks
________________
- Graph theory and graph algorithms; basic network measures

- Random networks and their characteristics (degree distribution, component sizes, clustering coefficient, network evolution), small world phenomena

- Scale-free property of networks, power-laws, hubs, universality

- Barabasi-Albert model, growth and preferential attachment, degree dynamics, diameter and clustering coefficient

- Evolving networks, Bianconi-Barabasi model, fitness, Bose-Einstein condensation

- Degree correlation, assortativity, degree correlations, structural cutoffs

- Network robustness, percolation theory, attack to
lerance, cascading failures

- Communities, modularity, community detection and evolution

- Spreading phenomena, epidemic modeling, contact networks, immunization, epidemic prediction

To give an application oriented introduction to modeling, optimization and control of mechanical systems with a focus on complex multibody systems in robotics and biomechanics.
To introduce computational tools to perform these tasks and apply them to examples in the computer exercises. While the lecture focuses on the theoretical and mathematical foundations of the field, the computer exercises serve to teach the usage of software tools for modeling, visualization, simulation and optimal control treating different example problems.
---------------------------------------------------------------
Dynamic process modeling
Mechanical basics, kinematics, dynamics
Multibody system modeling
Simulation of motions
Nonlinear optimization
Direct methods for optimal control problems
Elementary control principles
Basics of system dynamics
Open-loop and closed loop control of motions
Modeling human-like walking and running motions
Modeling locomotion of humanoid and bipedall robots
Stability of motions
----------------------------------------------------------
Prerequisites:
Programing skills in C/C++; basic knowledge in numerical analysis

Content:
- Introduction
- Perception and Color
- Raytracing
- Transformations
- Rasterization
- OpenGL
- Textures
- Curves
- Spatial Data Structures

Objectives:
The students understand fundamental and advanced concepts of computer graphics. They understand the mathematical fundamentals, data structures, and implementation aspects. They get to know raster graphics, geometric transforms, color perception and color models, and basics of geometric modeling. The students are able to apply these concepts to real-world problems using existing software packages, and develop small programs using OpenGL 4.

Prerequisites:
Acquiring basic knowledge in Computer Graphics.

Learn the full spectrum of algorithms for finding good approximate solutions to difficult combinatorial optimization problems or for solving them to proven optimality

Leaning Outcomes:
The participants
- can classify easy and hard problems,
- known the full spectrum of algorithms in combinatorial optimization,
- are able to model and solve difficult application problems with appropriate algorithms.
----------------------------------------------------
Lecture 4 h + Exercise course 2 h
- NP-hard optimization problems
- Approximative algorithms and heuristics
- Relaxations
- Branch-and-bound algorithms
- Dynamic programming
- Integer programming
- Polyhedral combinatorics
- Column generation and decomposition
- Traveling salesman problem
- Max-cut problem
-------------------------------------------------------
IP, IPK, IAD, MA4

In this tutorial we present an asynchronous data flow programming model for Partitioned Global Address Spaces (PGAS) as an alternative to the
programming model of MPI. GASPI, which stands for Global Address Space Programming Interface, is a partitioned global address space (PGAS) API.
The GASPI API is designed as a C/C++/Fortran library and focused on three key objectives: scalability, flexibility and fault tolerance. In
order to achieve its much improved scaling behaviour GASPI aims at asynchronous dataflow with remote completion, rather than
bulk-synchronous message exchanges. GASPI follows a single/multiple
program multiple data (SPMD/MPMD) approach and offers a small, yet
powerful API (see also www.gaspi.de and www.gpi-site.com). GASPI is successfully used in academic and industrial simulation applications.
Hands-on sessions (in C and Fortran) will allow users to immediately test and understand the basic constructs of GASPI. This course provides
scientific training in Computational Science, and in addition, the scientific exchange of the participants among themselves.

Please note that for the excercises the user needs to bring an own laptop and to already have access to either the bwUniCluster [1] oder the bwForCluster MLS&WISO Production [2] in Heidelberg/Mannheim. Please notify us in case of any problems. The course language is German (the slides are in English) or English if required.

Additional prerequisites for this course, the agenda and further information can be found at:
https://training.bwhpc.de/ilias/goto.php?target=crs_234&client_id=bwhpc

Please register by sending an e-mail to
hpc-support@urz.uni-heidelberg.de before November 21.

Our interviewee on this occassion is Prof. Dr. Hans-Peter Meinzer, a pioneer in the field of uniting medical- and information technology. Maria Ruppecht and Marcel Gutsche (both HGS MathComp) will ask him questions on his scientific career, work-life-balance and how to handle the workload in bio- and medical informatics within the university and industrial systems.

Heidelberg Chapter of SIAM / Public Lecture

Bitcoin´s blockchain emerges as an innovative tool which proves to be useful in a number of application scenarios. This talk will overview the security provisions of Bitcoin and its underlying blockchain-effectively capturing recently reported attacks and threats in the system. The talk will also discuss the limits of decentralization in Bitcoin´s blockchain and will outline a number of (open) challenges that should be overcome prior to large scale industrial deployment of open blockchains.

Dr. Karame is the manager and chief researcher of the security group at NEC Laboratories Europe.

He received his master of science in information networking from Carnegie Mellon University (CMU) in December 2006, and his Ph.D. degree in computer science from ETH Zurich, Switzerland, in 2011. Between 2011 and 2012, he worked as a postdoctoral researcher at the Institute of Information Security of ETH Zurich. Dr. Karame is interested in all aspects of security and privacy with a focus on cloud and blochchain security. Dr. Karame has recently co-authored a book on Bitcoin and Blockchain Security and has served on the program committees of a number of prestigious computer security conferences. More information about Dr. Karame can be found at ghassankarame.com.

Gene expression occurs in an environment in which both stochastic and diffusive effects are significant. Spatial stochastic simulations are computationally expensive compared to their deterministic counterparts and hence little is currently known of the significance of intrinsic noise in a spatial setting. I will show how starting from the reaction-diffusion master equation (RDME) describing stochastic reaction-diffusion processes, we can derive closed-form expressions for the approximate steady-state mean concentrations which are explicit functions of the dimensionality of space, rate constants and diffusion coefficients. These are generally different from those given by the deterministic theory of reaction-diffusion processes, thus highlighting the importance of intrinsic noise. Our theory is confirmed by comparison with stochastic simulations, using the RDME and Brownian dynamics, of two models of stochastic and spatial gene expression in single cells and tissues. Lastly, time permitting, I will discuss how one can extend these results to stochastic spatial simulations of intracellular processes which take into account macromolecular crowding, namely the volume exclusion due to the finite size of molecules.

Discussion with the speaker after the talk (at around 12:00), coffee will be provided.

The summer school is jointly organized by the IWR, Universität Heidelberg and Ateneo de Manila, Manila, Philippines. The school brings together experts from Germany and the Philippines to lecture on various topics of Scientific Computing

Machine Learning is the science and art of extracting meaningful information from data. In supervised machine learning, we use a training set of annotated examples to teach a computer to make valid predictions, or to reliably recognize items. In unsupervised learning, we use computers in an exploratory fashion, to discover interesting patterns in data.

Machine learning and pattern recognition are currently taking center stage in applications ranging from autonomous driving to social network analysis and drug development.

This seminar covers the essential techniques typically taught in an introductory lecture, including

statistical learning theory
generative and discriminative classifiers
ridge regression, lasso
logistic regression, generalized linear models
kernel methods, support vector machine
perceptron, multi-layer perceptron, neural networks
dimension reduction
cluster analysis
------------------------------------------
If you join, you will conduct a literature search on your topic, give a 45 min talk and summarize its contents in a report. You will receive 6 ECTS points and a grade based on: content of your talk (1/3), presentation (1/3) and quality of your report (1/3). This is a "Pflichtseminar" that is eligible towards the specialization in Computational Physics.

Deductive logic is the tool of choice when premises are or are not true. This seminar will focus on a generalization of logic when the aim is to maximize our expected utility in cases where a premise holds only with a certain probability.

In other words, we will study how to make optimal decisions under uncertainty. This relates to important problems in reinforcement learning, control theory and game theory.
-----------------------------------------------------------------
If you join, you will conduct a literature search on your topic, give a 45 min talk and summarize its contents in a report. You will receive 6 ECTS points and a grade based on: content of your talk (1/3), presentation (1/3) and quality of your report (1/3). This is a "Pflichtseminar" that is eligible towards the specialization in Computational Physics.

Kenntnis von Techniken des wissenschaftlichen Schreibens (insbesondere auch Literaturrecherche) Fähigkeit, komplexe wissenschaftliche Literatur zu erschließen
Erweiterte Fähigkeit, komplexe wissenschaftliche Literatur in einem Vortrag zu präsentieren
Erweiterte Fähigkeit, zu Vorträgen zu diskutieren und Feedback zu geben
Fähigkeit, ein kurze wissenschaftliche Ausarbeitung zu einem komplexen Thema zu erstellen
---------------------------------------------------------
Einführung in und Einübung von Techniken des wissenschaftlichen Schreibens
Vertiefte Einübung der Erschließung und Präsentation wissenschaftlicher Literatur
Fortgeschritteneres Informatikthema

Your goal is to learn how to develop a scientific project into an innovation. Along the question of how to become an Entrepreneur we enlarge your abilities in project-management, presentation and handling unknown situations.

Friday, 04.11. 09:00 to 17:00
• Human Centered Innovation/Design Thinking
• Creative Approach to find business aspects in your scientific work
• Design your first business model
• Test with users

Saturday, 05.11. 9:00 Uhr bis 17:00 Uhr.
• Question your business model
• Calculate your costs
• Prepare a presentation
• Present your results

Prerequisites: The course is open for all actual or former members of MathComp and the HGGS who are interested in getting new skills.
No prior knowledge is required.

Please register here

Registration deadline October 20, 2016

Manche Studien behaupten: Der Steuerspartrieb der Deutschen ist stärker ausgeprägt als der Sexualtrieb. Nun haben Studenten im Studium nur selten Steuerabgaben zu leisten, aber
dafür eine Vielzahl an Studienkosten (z.B. für den Laptop, für Lehrmaterialien, Fahrtkosten,
Telefon, Miete u.v.m.). Und diese Ausgaben können vom Finanzamt zurückgeholt werden,
spätestens ab dem Zeitpunkt in dem Steuern anfallen. Erfahrungsgemäß bedeutet das im
ersten Berufsjahr eine Steuererstattung in min. 4-stelliger Höhe. Dieses Geld dient vielen
Jobstartern für den Kauf einer Küche, einem Auto, der Bafög-Rückzahlung oder dem Erwerb
von Möbeln für die neue Wohnung.
Darüber hinaus gibt es auch wirtschaftliche Themen, die im Studium / Promotion bewegt
werden sollten und die eine immer wiederkehrende Steuererstattung ermöglichen. Dies wird
ebenfalls aufgezeigt.
Nach dem Training werden die Teilnehmer:
§ die Vorgehensweise kennen, wie sie ihre Studien- / Promotionskosten rückerstattet
bekommen.
§ für sich wichtige Themen erkennen und einen der größten Steuerhebel nutzen
können.
§ wissen, wie sie einen Verlustvortrag generieren & Werbungskosten und
Sonderausgaben steueroptimiert einordnen können.
Rahmendaten

Bitte hier anmelden

In this tutorial we present an asynchronous data flow programming model for Partitioned Global Address Spaces (PGAS) as an alternative to the programming model of MPI.
Hands-on sessions (in C and Fortran) will allow users to immediately test and understand the basic constructs of GASPI. This course provides scientific training in Computational Science, and in addition, the scientific exchange of the participants among themselves.

For more information and registration, please visit: https://training.bwhpc.de/ilias/goto.php?target=crs_234&client_id=bwhpc
--------------------------------------------------------
1. Basic knowledge in Unix / C or Fortran.
2. Bring along a laptop with wireless access via eduroam.
3. You need to have access to the bwUniCluster or the bwForCluster MLS&WISO in Heidelberg/Mannheim. Please note that having access to a bwForCluster other than MLS&WISO is not sufficient!
--------------------------------------------------------
http://www.gaspi.de/
--------------------------------------------------------
Learn to think and program highly parallel.

The talk introduces a special class of Boolean functions known as root functions. We show that there exists such class of Boolean functions, which can never appear as faulty response in irredundant two-level AND-OR combinational circuits even when any arbitrary multiple stuck-at faults are injected. Conversely, we show that any other Boolean function can appear as a faulty response in an irredundant realization of some root function under certain stuck-at faults. We characterize this new class of functions and explore their different properties. We show that for n variables, their number is exactly equal to the number of independent dominating sets in a Boolean n-cube. Also, we show that an n-variable root-function with minimum number of true minterms, where n = 2k - 1; k > 2, possesses some inherent error detection/correcting properties.

Dr. Debesh K. Das received the B.E. and M.E. degrees in electronics and telecommunication engineering and the Ph.D. degree in Engineering from Jadavpur University, Calcutta, India. He is currently with the Department of Computer Science and Engineering, Jadavpur University, as a Professor. Prior to this, he served as Minister-in-Charge of Information Technology in the Govt. of West Bengal. His research interests include logic synthesis and testing of VLSI circuits and fault-tolerant computing. He has published more than 100 papers in reputed Journals and International Conference Proceedings.

The unique electronic, optical, and mechanical properties of 2D materials have sparked an extraordinary level of theoretical and experimental activity. Stacking a few layers of 2D materials such as graphene and molybdenum disulfide, for example, opens the possibility to tune the electronic and optical properties of these materials. One of the main issues encountered in the mathematical and computational modeling of layered 2D materials is that lattice mismatch and rotations between the layers destroy the periodic character of the system.

Even basic concepts like the Cauchy-Born strain energy density, the electronic density of states, and the Kubo-Greenwood formulas for transport properties have not been given a rigorous analysis in the incommensurate setting. New approximate approaches will be discussed and the validity and efficiency of these approximations will be examined from mathematical and numerical analysis perspectives.

We show how the theory of measure-valued solutions can be used to show convergence for certain numerical schemes approximating the compressible Navier-Stokes system. Analysis is based on estimates based on the relative energy inequality adapted to oscillatory solutions.

After the lecture (4pm) you are cordially invited to "Meet the lecturer" and have a coffee and pretzel in the Common Room, 5th floor.

Dear colleagues,

It is our pleasure to announce a workshop on "Inverse problems: theory and statistical inference“ covering
- Variational methods for inverse problems
- Inverse problems in econometrics
- Posterior concentration in Bayes inverse problems
- Mathematical statistics and inverse problems

that will be held on Friday, October 28 and Saturday, October 29, 2016 at the International Academic Forum (IWH), Hauptstraße 242, Heidelberg.

The Workshop is Organized by MAThematics Center Heidelberg (MATCH) and Research Training Group "Statistical Modeling of Complex Systems and Processes“, Heidelberg/Mannheim.

The following speakers have kindly accepted our invitation:

Christoph Breunig (Humboldt-Universität zu Berlin)
Christina Butucea (Université Paris-Est Marne-la-Vallée) Marine Carrasco (Université de Montreal)
Fabienne Comte (Université Paris Descartes)
Thorsten Hohage (Georg-August-Universität Göttingen) Joel Horowitz (Northwestern University) Yuri Golubev (Aix-Marseille Université)
Christine de Mol (Université Libre de Bruxellles) Bartek Knapik (Vrije Universiteit Amsterdam)
Elena Resmerita (Alpen-Adria-Universität Klagenfurt)
Otmar Scherzer (Universität Wien)
Thomas Schuster (Universität des Saarlandes)
Anna Simoni (CNRS and CREST) Aad van der Vaart (Leiden University)

A poster session will complete the program. There are still open slots in the poster session, do not hesitate to ask your colleagues/students to submit an abstract or
paper to Sanja Juric .

There is no participation fee. Please announce your participation by email to Sanja Juric , but please note that the number of participants is limited and registration is on a first come first serve basis.

If you like to attend - the reception in the evening on Thursday, Oktober 27th and/or - the workshop dinner on Friday, October 28th, please contact Sanja Juric , she will provide further details.
Regularly updated information about the workshop can be found under: https://goo.gl/k7GcW7

Please note, there is change of time and location for the introductory course:
"An Introduction to Inverse and Ill-Posed Problems: Theory - Numerics - Applications“ by Professor Thomas Schuster (Universität des Saarlandes)
on Thursday, October, 27th, 14:15 - 15:45, Seminar room 7, 4th floor, MATHEMATIKON
16:15 - 17:45, Seminar room 2, 2nd floor, M?THEM?TIKON

The introductory course

“Nonparametric Instrumental Variables Estimation“ by Professor Joel Horowitz (Northwestern University)
will be held as announced earlier
on Tuesday, October, 25th, 9:15 - 10:40; 11:00 - 12:25; 14:00 - 15:25; 15:45 - 17:10 MATHEMATIKON, conference room, 5th floor

Dimension reduction or feature selection is an effective strategy to handle contaminated data and to deal with high dimensionality while providing better prediction. To deal with outlier proneness and spurious variables, we propose a method performing the outright rejection of discordant observations together with the selection of relevant variables.
To solve this problem, it is recasted as a mixed integer program which allows the use of efficient commercial solver. Also we propose an alternate projected gradient algorithm (proximal) so get a nice appoximated solution.

Given a polyhedral terrain and two points p, q on the terrain, a path joining p to q on the terrain is descending if the z-coordinate of a point v never increases while we move v along the path from p to q. The problem of finding shortest descending paths joining two given points on a polyhedral terrain was posed first by de Berg and van Kreveld (in Algorithmica, 18 (1997), pp. 306-322). Recently, the problem of finding approximations of such paths is considered by Ahmed, Lodha, and Lubiw (2010), Wei and Joneja (2013), Cheng and Jin (2014), etc. using Steiner point, graph, and sequence tree techniques.

In this talk we introduce the method of multiple shooting for solving the problem. It includes the factors: (f1) partition of the terrain, (f2) the straightness condition for the shortest descending paths at shooting points, and (f3) update of shooting points. In particular, the method does not rely on Steiner point, graph and sequence tree techniques on the entire terrain. If the straightness condition is satisfied then we obtain a local shortest descending path.

The corresponding algorithm is implemented in C++ using CGAL. Numerical results nonetheless indicate that the result is close to a local shortest descending path, even if it does not fulfil the straightness condition.

Interior point methods are well-suited for solving multistage stochastic
NLPs when an efficient algorithm for the huge structured KKT systems is
available. This is the case for large scenario trees with a moderate
number of variables per node: we present a distributed ``tree-sparse__
solution algorithm based on a static partitioning of the tree and
featuring low memory and communication overheads. We also address
structured quasi-Newton updates for the sparse Hessian as well as
structured inertia corrections to address non-convexity or
rank-deficiency of the KKT system.
Computational results for benchmark problems from portfolio optimization
and robust model predictive control demonstrate the performance of our
approach.

Mathematical models of tumor dynamics have become more accurate and accepted in recent years and enable a better prediction of initiation and development of a tumor, as well as its response to therapy. Using models one can compare different approaches or design new treatment strategies, which then can be tailored to individual patient data.

I will present some simple mathematical models for different problems in oncology; preventive gastric cancer screening, head and neck cancer response to radiation therapy, and combination therapies for pancreatic cancer. Mathematical models will be fit to retrospective clinical training data to derive parameter distributions for each participating mechanism. Parameter distributions with small variation will be collapsed into uniform rate constants, leaving variable mechanisms that are most likely to determine patient-specific outcomes. Calibrated models will be validated on independent training data, before virtual "in silico" trials determine optimal treatment protocols on a per patient basis.

Additional information:
After the talk there will be a meetup with the speaker to discuss different subjects in the area of cancer modelling. Discussion topics will include the impact of mathematical and computational models in the clinic, the challenges of interdisciplinary collaboration, and the current hot research topics. The meetup is tailored to provide a relaxed atmosphere where graduate students can have an introspect discussion of the field. Dr. Enderling will also address questions such as:
- What would you have wanted to know when you started working in this field?
- What are the "dos" and "don_ts" for PhD students starting their research in this field?
Master and PhD students are highly encouraged to attend.

Discussion with the speaker after the talk (at around 3pm) Common Room 5/303

The flows in close proximity to the air-sea interface control the exchanges between the atmosphere and the ocean, such as gases, heat, vapor, and momentum. The scales of the flow motions, including surface waves and turbulence, are much smaller than that of the atmosphere and the ocean. The processes within this surface layer, however, can profoundly affect the large-scale geophysical flows. The topic has been studied mainly by experimental measurements in the field and laboratory. Numerical simulation provides a complementary tool to reveal the underlying physics, but also pose challenges to the computations due to the inherent nonlinearity of the free boundary and the distinct length/time scales in the turbulent flow and wave motions. In this talk, I will briefly review the numerical method we have developed to solve this free-boundary problem. I will then present recent progress in evincing new physics in microscale air-sea interaction using numerical simulation. These studies are either aimed to elucidate previously unexplained observations or motivated/inspired by the recent innovative measurements.

To fully utilise computational resources for applications based on Discontinuous Galerkin discretisations, efficient and algorithmically optimal implementations are necessary. Traditionally a given PDE is solved by assembling a system of sparse equations and inverting the resulting matrix equation algebraically, for example with an AMG method. However, on modern multicore chip architectures with a poor FLOP-to-bandwidth ratio this approach becomes very expensive. Let N denote the number of unknowns per element. To apply the operator, in each grid cell a matrix of size N x N has to be loaded from memory and a dense matrix-vector multiplication with low arithmetic intensity is carried out; the overall cost of the method is O(N^2). This cost is reduced by matrix-free implementations where the matrix is re-calculated on-the fly. For tensor-product elements in d dimensions sum factorisation techniques reduce the computational complexity from O(N^2)=O(n^{2d}) to O(d*n^{d+1}), where n is the number of unknowns in one direction.

In preconditioned Krylov-subspace solvers and multigrid smoothers it is often necessary to invert block-diagonal matrices. Even though the action of A is implemented in a matrix-free way, local block-matrices D_e of size n^d x n^d are assembled in each cell and then inverted with an exact LU- or Cholesky-factorisation. Overall this requires O(n^{2d}) bandwidth-bound operations and quickly becomes the bottleneck of the solver as the order n increases. To circumvent this problem, we solve the system D_e.x=y approximately with an iterative method. Since the application of D_e can be implemented in a matrix-free way, the action of D_e^{-1} becomes FLOP bound and the cost decreases from O(n^{2d}) to O(n_{iter} * d * n^{d+1}) where n_{iter} is the number of iterations required to solve the system in each cell.

We study the efficiency of this approach for the solution of linear convection-diffusion systems; problems of this type arise, for example, in operator splitting approaches for unstable porous media flow. We demonstrate the algorithmic and computational efficiency of the method for a hybrid multigrid algorithm with hp-coarsening, similar to [Bastian et al. (2012), Num. Lin. Alg. with Appl. 19 (2), pp. 367-388]: on the finest level a matrix-free block-Jacobi or block-SSOR smoother is applied to the high-order system, and the low-order system on the coarser levels is solved with AMG.

All code is implemented in the EXADUNE code base and we demonstrate the efficiency of our approach for a range of elliptic PDEs, including a convection-dominated problem and the stationary SPE10 benchmark.

Authors: Eike Müller (University of Bath), Peter Bastian, Steffen Müthing, Marian Piatkowski (Heidelberg University)

tba

Die Bedeutung der Mentalen Leistungsfähigkeit nimmt im Wandel der kognitiven Anforderungen einer modernen Arbeitswelt stetig zu. Der Zusammenhang von psychomentaler Überlastung und kognitiver Einschränkungen wird häufig unterschätzt, besonders bei Menschen mit hohem Leistungspotenzial bzw. im Rahmen anspruchsvoller beruflicher Tätigkeiten.
Es wird im Vortrag dargestellt, wie die kognitive Leistung gefördert werden kann sowie die Folgen stressbezogener Belastung minimiert werden können.
In der Einführung werden, je nach Vorkenntnissen der Teilnehmer, theoretische Aspekte für das Verständnis des Themas vorgestellt. Es werden ebenso neuere wissenschaftliche Erkenntnisse im Rahmen des Vortrags berücksichtigt werden.
Zwei Techniken aus der Hypnotherapie werden praktisch eingeübt. Sollte es der zeitliche Rahmen erlauben, können die vorgestellten Übungen an den persönlichen Bedürfnissen angepasst werden.
Praxis und Theorie werden in 90 Minuten inhaltlich auf einander abgestimmt.

Flip-Chart-Arbeit - Power-Point - praktische Übungen - Diskussion und Fragen

Bitte hier registrieren

The short course addresses the improvement of medical health with respect to the treatment of the human respiratory system as well as that of cerebral aneurysms with flow diverter stents. Different views from representatives of the Medical Schools as well as experts in computational fluid dynamics and from aerosol devices enlighten the problems and their solutions with new treatment options.