HGS MathComp Curriculum & Events

The talk is devoted to modeling, optimization methods and related tools for decision support in software release planning. Release decisions are
part of product management, an emerging discipline that aims at improving the whole process of development ranging from product requirements to design, implementation, product launch, and service. In its essence, release planning is the process of defining the functionality of a sequence of product releases as part of incremental development. Customers and stakeholders are continuously asking for more features or revision of existing ones. But: Which ones to finally select
for the next release and why? And: Which features are not attractive enough and are better left out (or postponed)? Release planning is a cognitively and computationally challenging problem because of its size, the number of stakeholders involved in it, the variety of variables that need to be looked at, and the uncertainty of the information it is relying upon. The overall goal is to determine a “most promising”
release plan such that the total degree of satisfaction of all the different stakeholders is maximized.

Dr. Ruhe (http://sern.cpsc.ucalgary.ca/~ruhe/) received a doctorate rer. nat degree (Applied Mathematics) from Freiberg University, Germany and a doctorate habil. nat. degree (Informatics) from University of Kaiserslautern, Germany. From 1996 until 2001 he was deputy director of the Fraunhofer Institute for Experimental Software Engineering Fh IESE. Since 2001, Ruhe holds an Industrial Research Chair in Software Engineering at University of Calgary. He has extensive experience in industry collaboration and technology transfer. He has organized different international conferences and workshops, has given keynotes at various conferences and has published more than 180 reviewed research papers at journals, workshops and conferences. In the area of software
release planning, Ruhe recently published a book and received a US patent. He is the Associate Editor of the Elsevier Journal Information and Software Technology IST and a member of the Editorial Board of various other journals (such as JSME and IJSEKE). He is a member of the ACM, the IEEE Computer Society, and the German Informatics Society GI.

Knowledge about project organization and processes, models, cost estimation, risk management, controlling, project roles, communication and interfaces to other tasks in IT projects.

Literature:
PMI (Project Management Institute): A Guide to the Project Management Body of Knowledge (PM BOK ® Guide), 4. Ausgabe 2008

Topics:

- why computational image? Limitations of tradiational imaging techniques
- Light fields, plenoptic function, radiation transport
- optical properties of surfaces: bidirectional reflection distribution function (BRDF)
- volumetric optical properties: absorption and scattering
- light fields and wave optics: Wigner distribution function
- Standard image acquisition techniques as sparse sampling of light fields
- light field simulation
- capturing of light fields
- camera arrays
- plenoptic cameras
- MTF engineering, extended depth of field
- depth edge detection by multiflash imaging
- spatially modulated light fields
- coded exposure cameras
- time-of-flight imaging
- super resolution

This course is targeted at non-mathematicians in the HGS and offers an introduction to tools like:
Latex+bibtex+JabRef+beamer+Sweave, gnuplot (for plotting functions), basic statistics with R plus a small course on “how to lie with statistics”, inkscape for drawing images, if wanted also basic HTML/CSS manipulation with Phase5, and blog writing with blogspot.com. At the first meeting we will collect the suggestions and needs of the audience and design the compact course accordingly.

Abstract:
The complex behavior of turbulence which involves the mixture of the wide range of length and time
scales is the consequence of a fairly simple set of equations – Navier-Stokes equations. However,
analytical solutions to even the simplest turbulent flows do not exist. A complete description of
turbulent flow can only be obtained by numerically solving Navier-Stokes equations. Instead of
solving the statistical evolution of the flow, or including the parameterization of small-scale
turbulence, direct numerical simulation (DNS) aims to resolve the instantaneous flow field. In the
first part of this lecture, we review some contribution of DNS to the turbulence research.
It is well-known that gas exchange across air-sea interface is dominated by the turbulent boundary
layer in the water side. To understand the turbulent behavior in this boundary layer, except to solve
the Navier-Stokes equations, the statistics-based, conceptual model, e.g. surface-renewal model,
random-eddy model, etc. and the hydrodynamics-based model, e.g. large-eddy model, small-eddy
model etc. have been accepted to describe the transfer processes across air-sea interface. In the
second part of the lecture, we briefly introduce these models and their application.
Contents:
1. Numerical solution for a three-dimensional, unsteady, incompressible free-surface flow
1.1 Unsteady boundary: a free-surface
1.2 Combined scheme of pseudo-spectral method and finite-different method
1.3 Modified Newton’s method for pressure Poisson equation
1.4 General comments for direct numerical simulation
2. Transport model of mass exchange across air-water surface
2.1 Free-surface turbulence
2.2 Conceptual models
2.3 Hydrodynamics-based models
2.4 Applications of the transport models


Reference:
1. Moin, P. and K. Mahesh (1998), Direct numerical simulation: A tool in turbulence research.
Annu. Rev. Fluid Mech., 30, 539-578.
2. Jähne, B. and H. Haußecker (1998), Air-water gas exchange. Annu. Rev. Fluid Mech., 30, 443-
468.

This course targets at students who have a first understanding of working with the R software (see web page for prerequisites -- I assess that they can be obtained by home study). In three sessions specific topics are tackled. We will put the course into the curriculum as well.

Working as a PhD student you have the challenging task of developing research findings and write you doctoral thesis within three years. This alone is a demanding job. In addition, it is vital to the scientific process that your findings are presented to the scientific community. For most PhD students this is the first big project in their professional life and it could have a crucial impact on their future professional career. PhD students are highly motivated when they start their PhD studies but may underestimate the need for professional management for this three-year project \"doctoral thesis\".

This seminar demonstrates how to approach the doctoral thesis in a professional way. Project management tools and techniques are used, tailored to the specific situation of PhD students. You will learn how to set a project vision, define clear objectives, gain buy-in from your supervisor and other colleagues in your group, and how to develop a project plan, which is structured and at the same time flexible enough to easily adjust to unexpected findings. You will establish a \"controlling cycle\" which helps you to recognise risks and problems as early as possible, and you will learn how to manage critical situations and deal with ups and downs. Furthermore, networking with colleagues, supervisors and other people is an important topic of this seminar.

Throughout the seminar, you will work on your own doctoral thesis and share your experience with others. This seminar is most beneficial for PhD students who are in the early phases of their doctoral thesis. At the end of the seminar you will have established a strategy on how to approach your own doctoral thesis. During the follow-up REVIEW we will share experience and best practices and deal with open questions from the first module.

This seminar will help you to make the most effective use of your three years and finish your doctoral thesis on time. You will also learn and practise the basic concepts of project management - which are required in industry and research institutions.

Teilnehmer erhalten wichtige Informationen zu verschiedenen Übungen, damit ein bevorstehendes Assessment Center (AC) für sie bedenkenlos und erfolgreich verläuft. Teilnehmer trainieren unter anderem Selbstpräsentation, Gruppendiskussion und andere relevante Aufgaben aus Gruppenauswahlverfahren. Darüberhinaus erhalten die teilnehmer Informationen zu den Erwartungen der Personalverantwortlichen. Weiter lernen sie, worauf besonders geachtet wird und wie Bewerber auftreten sollen. Die teilnehmer erhalten ein validiertes Feedback zu ihren rethorischen Präsentationen anhand einer Videoanalyse.


Please register here:

http://hgs.iwr.uni-heidelberg.de/Portfolio_HGS/VERANSTALTUNGEN/reg_form/reg_form.php?id=61

Preliminary discussion 17.10.2011, 2.15 pm.
Objectives and content:
To have a firm command of quantum chemistry-based modeling of molecular structures and properties Acquiring basic knowledge in HF, post HF, DFT theories. Advanced studies based on hands-on exercises fundamental knowledge of chemistry and physics.

Die Vorlesung gibt eine Einführung in die Theorie und Numerik von Optimierungsproblemen, bei denen partielle Differentialgleichungen (PDE) als Nebenbedingungen auftreten.

Die Vorlesung beschäftigt sich mit

1) Parameterschätzung: Abschätzung von Parametern bei elliptischen und parabolischen PDE.

2) Schlecht gestellte inverse Probleme: Abschätzung verteilter Parameter bei elliptischen PDE.

3) Optimale Versuchsplanung mit PDE: Abschätzung und Optimierung der Kovarianz der Parameter.

4) Optimale Steuerung von PDE.

Literatur:

1) H. Banks, K. Kunisch: Estimation Techniques for Distributed Parameter Systems, Birkhauser Boston Inc, (1989)

2) H.W.Engl, M.Hanke, A.Neubauer, Regularization of Inverse Problems, Kluwer, 2008

3) M. Hinze, R. Pinnau, M. Ulbrich, S. Ulbrich, Optimization with PDE Constraints, Springer, 2008.

4) K. Ito, K. Kunisch, Lagrange multiplier approach to variational problems and applications, Society for
Industrial and Applied Mathematics, 2008

5) D.G. Luenberger, Optimization by Vector Space Methods, Wiley, 1998

6) D.G. Luenberger, Linear and Nonlinear Programming, Springer, Berlin, 2008

7) F. Tröltzsch, Optimale Steuerung partieller Differentialgleichungen, Vieweg, 2009

8) D. Ucinski, Optimal Measurement Methods for Distributed Parameter System Identification, Crc Pr Inc (2005)

Zielgruppe:

Studierende der Mathematik, Informatik und Physik im Hauptstudium oder Promotionsstudium. .

Voraussetzungen:

Stoff der Vorlesung Numerik Partieller Differentialgleichungen ist wünschenswert. Kenntnisse des Stoffes aus Numerische Optimierung bei Differentialgleichungen sind hilfreich. Wünschenswert aber nicht notwendig sind Kenntnisse der Funktionalanalysis.

The course and lecture introduces time-discrete and time-continous mathematical models in form of the difference equations and ordinary differential equations and presents their applications in biology. We provide mathematical methods to study dynamics of such models, existence and linear stability of the the steady states and qualitative analysis of time-dependent dynamics. For this purpose we introduce basic knowledge of complex numbers, systems of linear algebraic equations, matrices, their eigenvalues and eigenvectors; recursive methods, and ordinary differential equations. The theory is illustrated by examples from mathematical biology.

Objectives:
To provide a background in mathematical methods of difference equations and ordinary differential equations for mathematical modelling of biological systems

Existing software has to be arranged for a specific issue and is tested by using test excerises. A self-made algorithm on basis of open source libraries and/or technical papers should expand a programs functionality. At the beginnig of next term we introduce current topics in a preliminary discussion. Your own topic suggestions are warmly werlcome! They will be discussed with your supervisor and fixed in an scheduled specification sheet. Your work on it will be on your own or max. in groups of two or three.

The content of the tutorial will be based on the R-package INLA
(http://www.r-inla.org/), where INLA stands for integrated nested
Laplace approximation, and the theory related to that software. This
includes Bayesian inference for data where (some of) the parameters
can be described through Gaussian Markov Random Fields (GMRFs) and the
connection between GMRFs and stochastic partial differential
equations.

The INLA software is largely based on two papers published in the
Journal of the Royal Statistical Society, Series B:
http://www.math.ntnu.no/~hrue/r-inla.org/papers/inla-rss.pdf
http://onlinelibrary.wiley.com/doi/10.1111/j.1467-9868.2011.00777.x/full

In order to get the most out of the tutorial, the attendees should
bring their own laptop with R and INLA installed.

Seit der Einführung des Computers als Forschungs-, Experimentier- und Prognoseinstrument erleben die Wissenschaften einen tief greifenden Wandel. Nicht nur die Praktiken und Infrastrukturen wissenschaftlichen
Arbeitens verändern sich, sondern auch die Logik der Forschung unterliegt einer grundlegenden Transformation. Neben Theorie, Experiment und Messung eröffnen Computerexperimente ein neues Feld der Wissensproduktion und verändern radikal die Experimentalkultur der Naturwissenschaften. Der Vortrag zeigt am Beispiel der Klimaforschung und der Biologie, welchen Einfluss die _in-silico_ Experimentalkultur hat.

In this joint work with Sourav Chatterjee we consider a dense random graph with a large
number N of vertices and the edges are independently on with probability p and off with
probability 1-p. There is a law of large numbers for the number of triangles as well as for number of various finite graphs that occur in the random graph. We derive a large deviation formula for these numbers, and predict which types of graphs are responsible for these large deviations. There is an interesting phase transition. The same method can also handle large deviations for eigenvalues of random symmetric matrices. A crucial ingredient
is Szemeredi’s regularity lemma and its consequences.

Ganzzahlige und kombinatorische Optimierung ist ein relativ junges Gebiet. Die ersten grundlegenden Arbeiten stammen aus den 50er und 60er Jahren des vorigen Jahrhunderts. Danach hat sich das Gebiet, nicht zuletzt auch durch die enorme Leistungssteigerung der Computer, stürmisch weiterentwickeltund stellt heute wesentliche Methoden zur Modellierung und Lösung praktischer Probleme bereit.
In diesem Seminar soll anhand des Buches \"50 Years of Integer Programming 1958-2008: From the Early Years to the State-of-the-Art\" ein Teil der ersten bahnbrechenden Arbeiten vorgestellt und aus heutiger Sicht diskutiert werden.
Zur erfolgreichen Teilnahme am Seminar sind ein Vortrag und eine schriftliche Ausarbeitung erforderlich. Die Teilnahme wird mit 4 ECTS-Leistungspunkten bescheinigt. Die Veranstaltung richtet sich an Studierende in Mathematik oder Informatik. Notwendige Voraussetzung zur Teilnahme ist die erfolgreiche Absolvierung der Module \"Effiziente Algorithmen 1\" oder \"Effiziente Algorithmen 2\".

This course is offered by the Statlab, heidelberg University

Topics covered:
* Basics of R programming
* Linear Models
* Diagnostics and tests for univariate distributions
* Empirical distribution functions and related statistics
* Residual analysis and regression diagnostics
* distribution-independent methods
* Permutation and resampling methods
* Multivariate problems
* Graphics for multivariate methods and simulations

On Friday, the 20th of January, I will hold my lecture on _How to lie with statistics_. If you do not know yet how to do that, you are heartily invited to join us! ;-) The lecture has three key points:

1) Statistics is sometimes not very intuitive,
2) those that do not engage themselves in the topic might easily fool
themselves or
3) ... be fooled by others.

We will speak about how to come to scientific conclusion (,,The scientific method´´ - very basic treatment), about the role of statistics in this endeavour, define _statistical significance_ and speak about its interpretations.

We then treat various paradoxa and (seemingly) unintuitive findings like the Yule-Simpson paradoxon or the Monty-Hall-Problem. Finally I will show various examples of where Statistics has gone awry - partly unintended, partly intended. The course wants to turn you into a literate reader (and producer) of basic statistics. We will also speak about how dead salmons can recognize human face expressions and why a positive HIV test or a positive indication for breast/prostrate cancers does not necessarily give a high probability that someone is actually infected/has cancer.

The lecture is intended for those of you which need to deal with understanding statitistics in literature or producing ones for your own work or just interested readers of a daily newspaper.

Join us in the CIP-Pool of the physical institute, Albert-Überle-Straße 5, 11 s.t.-12:30 and 13:00-14:30.

>>>>> There are approx 10 places left, so please let me know whether you want to come (katharina.zweig@iwr.uni-heidelberg.de). <<<<<<

If you know some fishy statistics from your own field of research, please bring it along!

This work presents a two-stage motion planner for walking humanoid robots. A first draft path is computed using random motion planning techniques that ensure collision avoidance. In a second step, the draft path is approximated by a whole-body dynamically stable walk trajectory. The contributions of this work are:
(i) a formal guarantee, based on small-space
controllability criteria, that the first draft path can be approximated by a collision-free dynamically stable trajectory;
(ii) an algorithm that uses this theoretical property to find a solution trajectory. We have applied our method on several problems where whole-body planning and walk are needed, and the results have been validated on a real platform: the robot HRP-2.

An algorithm for solving continuous-time stochastic optimal control problems is presented. The numerical scheme is based on the stochastic maximum principle (SMP) as an alternative to the widely studied dynamic programming principle (DDP). By using the SMP, Peng 1990 obtained a system of coupled forward-backward stochastic differential equations (FBSDE) with an external optimality condition. We extend a known numerical scheme by a Newton-Raphson method to solve the FBSDE system and the optimality condition simultaneously. As far as the authors are aware, this is the first fully explicit numerical scheme for the solution of optimal control problems through the solution of the corresponding extended FBSDE system. We discuss possible numerical advantages to the DDP approach and consider an optimal investment-consumption problem as an example.

This talk has two distinct, but closely related sections: in the first, we first discuss the basic differences between mainstream computer graphics, and predictive image synthesis, and also point out
the creative possibilities that this form of graphics can offer. In the second part, we give a brief overview of the application domains predictive rendering is useful for, the technological state of the
art in this field, and the main research directions that are currently being investigated. This includes the specific topics that our group in Prague is working on now, and which directions will probably be upcoming research areas in the near term future.

Immense amounts of high resolution data are now routinely produced thanks to recent advances in imaging techniques. State-of-the-art Computer Vision algorithms designed to operate on natural 2D images tend to perform poorly when applied to b microscopy image stacks for a number of reasons. The size of a typical image stack renders many segmentation schemes intractable. Most approaches rely on local statistics that easily become confused when confronted with the noise and textures. We have proposed an automated graph partitioning scheme that addresses these issues. It reduces the computational complexity by operating on supervoxels instead of voxels, incorporates global shape features capable of describing the 3D shape of the target objects, and learns to recognize the distinctive appearance of true boundaries. Our approach banks on the SLIC superpixels frameworks which has been demonstrated to be a very fast algorithm to create compact and boundary preserving superpixels.

One of the greatest scientific advances of the 20th Century was not
substantive, but methodological: the laying out by Fisher of the
principles of sound experimentation, so allowing valid conclusions to be
drawn about the effects of interventions - what we must surely regard as
\\\\\\\\\\\\\\\"causal inference\\\\\\\\\\\\\\\". More recently \\\\\\\\\\\\\\\"causal inference\\\\\\\\\\\\\\\" has developed as a
major enterprise in its own right, with its own specialist formulations
and methods; however, these owe more to Neyman than to Fisher. In this
talk I shall explore the connexions and contrasts between older and
newer ideas in causal inference, revisit an old argument between Neyman
and Fisher, and argue for the restructuring of modern theories of causal
inference along more Fisherian lines.

The local and superlinear convergence of the generalized Newton_s method on regular semi-smooth problems established by Kummer, Qi and others has generated a lot of interest in the last two decades. However theory and algorithmic realization suffer from two related difficulties. Firstly the domain of convergence is unstable with respect to arbitrarily small perturbations of the right hand side. Secondly, this problem of miniscule domains of contraction cannot be overcome by simple stabilisations like line-searches or trust regions. Thirdly, the differentiation rules for generalized Jacobians are only inclusions so that guaranteed elements of these multifunctions can only be produced in exceptional cases. Finally, the considerable effort needed to set up the computation of generalized Jacobians at points of nondifferentiability is usually wasted, because all
actual iterates generated by a solver or optimizer are points of differentiability with probability one. We will indicate how all four problems can be overcome by algorithmic piecewise linearization.