
Mathematics of Data Analysis  Oberseminar
News Dates Seminar Literature Contact
News
top
 18.10.2016: We restart the seminar with a talk of Michael Sandbichler.
Dates
top
Tuesday, 18.07.2017 
14:3016:00 
Speaker: 
Prof. Uwe Kähler ^{} (Universidad de Aveiro, Portugal) 
Title: 
TBA 
Room: 
02.08.020 
Tuesday, 27.06.2017 
14:3016:00 
Speaker: 
Prof. Ahmed I. Zayed ^{} ( DePaul University, Chicago) 
Title: 
TBA 
Room: 
02.08.020 
Tuesday, 21.06.2017 
10:3011:30 
Speaker: 
Timo Klock ^{} (Simula Research Laboratory) 
Title: 
Inverse problems of unmixing type arise in many reallife applications such as audio processing or medical image analysis. In such problems additive noise directly affects a sparse signal before being measured through a sampling matrix. Consequently, the noise in the measurement is amplified through the sampling process and the socalled noise folding phenomenon occurs. This amplification worsens the results on support identification by means of "classical'' sparse recovery techniques based on the l1penalised Lasso functional. Several recent works suggest to apply a multipenalty framework for a correct modeling and separation of the original signal in such problems. Admittedly, the parameter choice in such multipenalty functionals becomes more involved compared other singlepenalty methods. In this talk, we use multipenalty regularization for the unmixing problem. It has been shown that the resulting functional can be seen as a parameterised Lasso such that we can solve the multipenalty minimisation with known techniques for the Lasso. In this spirit, we provide an extension of the Lassopath algorithm for an efficient calculation of large parts of the multipenalty solution space without performing extensive gridsearches over regularisation parameters. Finally, by using a naive support selection heuristic based on signaltonoise ratios, we identify a unique support that is compared to results from conventional sparse recovery techniques. Such experiments confirm improved results of the multipenalty functional compared to singlepenalty counterparts. 
Room: 
TBA 
Tuesday, 06.06.2017 
14:3016:00 
Speaker: 
Prof. Stephan Günnemann (TUM) 
Title: 
Efficient and Robust Learning with Graphs 
Room: 
02.08.020 
Tuesday, 30.05.2017 
14:3016:00 
Speaker: 
Prof. Felix Krahmer (TUM) 
Title: 
In this talk, we will discuss random models with varying degrees of imposed structure for different applications in signal processing and data analysis. First, we will study a matrix factorization problem as motivated by applications in bioinformatics. To establish uniqueness under a random model, we develop new tools in probabilistic combinatorics. Motivated by applications in wireless communication, we consider the problem of simultaneous demixing and deconvolution for randomly embedded signals. We improve upon recent results by Ling and Strohmer, establishing for the first time nearoptimal parameter dependence. Lastly, we show nearoptimal recovery guarantees for analogtodigital conversion in combination with compressed sensing for structured random measurement systems. These are joint works with the speaker’s PhD students David James, Dominik Stöger, and JoeMei Feng as well as with Matthias Hein (Universität des Saarlandes), Peter Jung (TU Berlin), and Rayan Saab (UC San Diego). 
Room: 
02.06.011 
Thursday, 18.05.2017 
09:0010:00 
Speaker: 
Matthias Beckmann ^{} (Universität Hamburg) 
Title: 
This talk concerns the approximation of bivariate functions by using the well established filtered back projection (FBP) formula from computerized tomography, which allows us to reconstruct a bivariate function from given Radon data. Our aim is to analyse the inherent FBP approximation error which is incurred by the application of a lowpass filter. To this end, we present error estimates in Sobolev spaces of fractional order. The obtained error bounds depend on the bandwidth of the utilized filter, on the flatness of the filter’s window function at the origin, on the smoothness of the target function, and on the order of the considered Sobolev norm. Finally, we prove convergence for the approximate FBP reconstruction in the treated Sobolev norms along with asymptotic convergence rates, as the filter’s bandwidth goes to infinity. The theoretical results are supported by numerical experiments. This talk is based on joint work with Armin Iske. 
Room: 
02.08.011 
Tuesday, 16.05.2017 
14:3016:00 
Speaker: 
Benedikt Diederichs (Universität Hamburg) 
Title: 
Prony's problem  estimating the frequencies of an exponential sum  and its higher dimensional analogs have attracted a lot of attention in recent years. A somewhat neglected question is whether this problem is wellposed. In this talk, some results in this direction will be presented. The most important techniques we need are e cient estimates of certain exponential sums. Inci dentally, they can be used to improve classic estimates of the condition numbers of matrices arising when one interpolates with a positive definite kernel. If time permits, we will discuss this connection. This talk is based on joint work with Armin Iske. 
Room: 
02.08.020 
Tuesday, 09.05.2017 
14:3016:00 
Speaker: 
Markus Stachl (TUM) 
Title: 
The flow of motorized vehicles through urban road networks is known to be one of the main reasons for high pollution in metropolitan areas. So far little scientific research has been spent on the effects of coordinated traffic lights on emissions. In our approach to simulate traffic flow through a network of roads we resort to a wellposed macroscopic conservation law coupled with a onedimensional pollution model. Model Predictive Control (MPC) is used as a responsive optimization technique to manage the movement of cars close to junctions, mirroring the use of traffic signals. On basis of an exemplaric road network in Munich we show that by optimizing traffic dynamics in this manner a decrease in COemissions by 510% can be achieved. 
Room: 
02.08.020 
Tuesday, 02.05.2017 
14:3016:00 
Speaker: 
Frank Filbir ^{} (Helmholtz Zentrum München) 
Title: 
We develop constructive algorithms to represent functions defined on a metric measure space within a prescribed accuracy. The constructions can be based on either spectral information or scattered samples of the target function. Our algorithmic scheme is asymptotically optimal in the sense of nonlinear nwidths and asymptotically optimal up to a logarithmic factor with respect to the metric entropy. The talk is based on joint work with Martin Ehler, University of Vienna. 
Room: 
02.08.020 
Tuesday, 04.04.2017 
10:3011:30 
Speaker: 
Thomas Fink ^{} (Universität Passau) 
Title: 
In various fields of image analysis, determining the precise geometry of occurent edges, e.g. the contour of an object, is a crucial task. Especially the curvature of an edge is of great practical relevance. In this talk, we introduce an extension of the continuous shearlet transform which additionally utilizes shears of higher order. This extension, called the Taylorlet transform, allows for a detection of the position and orientation, as well as the curvature and other higher order geometric information of edges. Furthermore, we introduce novel vanishing moment conditions of the form \( \int_{\mathbb{R}} g \left( \pm t^k \right) t^m dt \). We will show that Taylorlets fulfilling such conditions enable a more robust detection of the geometric edge features. 
Room: 
02.10.011 
Tuesday, 28.03.2017 
10:3011:30 
Speaker: 
Dr. Ilaria Lucardesi ^{} (École des Mines de Nancy) 
Title: 
The two most studied elliptic PDEs are probably the torsion problem, also known as StVenant problem, and the Dirichlet eigenvalue problem. For these classical problems, many estimates and qualitative properties have been obtained, see for example works by P\'olya, Szeg\"o, Schiffer, Payne, Hersch, Bandle, and many others. In this seminar I present some recent results about upper and lower bounds of two shape functionals involving the maximum of the torsion function: I consider the ratio $T(Omega)\lambda_1(Omega) Omega$ and the product $M(Omega)\lambda_1(Omega)$, where $Omega$ is bounded open set with finite Lebesgue measure $Omega$, $T(\Omega)$ denotes the torsion, and $\lambda_1(\Omega)$ the first Dirichlet eigenvalue. Particular attention is devoted to the subclass of convex sets. 
Room: 
02.12.020 
Wednesday, 01.03.2017 
11:0012:00 
Speaker: 
Patrick van Meurs ^{} (Kanazawa University) 
Title: 
The starting point is a 2D model for the dynamics of n dislocations, which are modelled as point particles with a positive or negative ’charge’. In the celebrated engineering paper by Groma and Balogh in 1999, the limit passage n → ∞ of these dislocation dynamics is performed in a statistical mechanics framework, which relies on a phenomenological closure assumption. In my talk, I present how to pass rigorously to the limit n → ∞ by using the theory of Wasserstein gradient flows and using advanced functional analysis on the weak form of the evolution equation. Interestingly, our conclusion for the limiting dynamics of the dislocation density differs from the conclusion in the paper by Groma and Balogh. 
Room: 
02.08.011 
Tuesday, 07.02.2017 
16:0017:00 
Speaker: 
Kiryung Lee ^{} (Georgia Tech) 
Title: 
Multichannel blind deconvolution resolves unknown input signal from multiple channel outputs with unknown impulse responses. It is often easier to estimate only the unknown channel impulse responses when they are modeled with only few parameters. This is the case with channel estimation problems in wireless communications and underwater acoustics. In particular with FIR models, various reconstruction methods based on statistics of the input signal and/or the commutativity of the convolution operator have been proposed with algebraic performance guarantees in 1990s. However, these guarantees are restricted to the case where the observations are noisefree or the input has infinite length. In fact, with finitely many noisy observations, the empirical performance of these classical methods deteriorates dramatically as the length of observation decreases. This weakness restricts their utility in estimating timevarying channels. Motivated from the observation that parametric models with few parameters can be embedded into lowdimensional subspaces, we propose modifications on the classical methods leveraging these subspace models. Our proposed method provides significant improvement over the original method in terms of providing stable estimates from finitely many noisy measurements. Furthermore, using recently developed tools in applied probability, we derive rigorous performance guarantees under certain random subspace models. We verify that the numerical results are aligned with our theory through Monte Carlo simulations. Furthermore, we also verify the empirical success in a realistic setup of channel delay estimation. Finally, we extend the methods and theory to the sparsity case and present corresponding theoretic and numerical results. This is the joint work with Felix Krahmer and Justin Romberg. 
Room: 
02.12.020 
Tuesday, 31.01.2017 
16:0017:00 
Speaker: 
Stefano Almi (TUM) 
Title: 
The aim of this talk is to present a variational model of the quasistatic evolution of hydraulic crack in the general framework of rateindependent processes. 
Room: 
02.06.020 
Tuesday, 24.01.2017 
16:0017:30 
Speaker: 
Marco Morandotti ^{} (TUM) 
Title: 
In this talk I will present the theory of structured deformations. Starting from the original formulation by Del Piero and Owen, I will move to the variational formulation proposed by Choksi and Fonseca, to conclude with some recent results regarding relaxation of nonconvex energies. I will also present some examples and explicit formulas as an application to specific surface energies. 
Room: 
02.06.020 
Tuesday, 17.01.2017 
10:3012:00 
Speaker: 
Giacomo Albi ^{} (TUM) 
Title: 
In this talk I will review some basic results in the modeling and control of multiagent system. The prototype problem will account a large system of interacting agents, whose dynamics are influenced by a policy maker acting in order to enforce a desired behavior. Different examples will be shown: from opinionformation processes to crowdsafety management. From the mathematical viewpoint, this situation can be described by means of a meanfield optimal control problem, governing the dynamics of the probability distribution of the agent population. In order to deal numerically with the highdimensionality and the nonlinearities of this type of problems, I will introduce a novel approximating hierarchy of suboptimal controls based on a stochasticBoltzmann approach, whose computation requires a moderate computational effort compared with standard direct approaches. I will compare the behavior of the control hierarchy with respect to the solution of the optimal control problem. Several numerical examples will show the effectiveness of the proposed strategies. 
Room: 
02.10.011 
Tuesday, 10.01.2017 
10:3012:00 
Speaker: 
Prof. Dr. Sergei Pereverzyev ^{} (RICAM Linz) 
Title: 
We discuss the parameter choice in learning algorithms generated by general regularization scheme. In contrast to classical deterministic regularization, the performance of regularized learning algorithms is influenced not only by the smoothness of a target function, but also by the capacity of a regularization space. In supervised learning both the smoothness and the capacity are intrinsically unknown. Therefore, we are interested in a posteriori regularization parameter choice rules and propose a new form of the balancing principle. We provide the analysis of the proposed rule and demonstrate its advantages in simulations. Joint research with Peter Mathe (WIASBerlin) and Shuai Lu (Fudan University, Shanghai). 
Room: 
02.10.011 
Tuesday, 20.12.2016 
11:0012:00 
Speaker: 
David James ^{} (Göttingen) 
Title: 
In this talk, we investigate subspaces spanned by biased random vectors. The underlying random model is motivated by applications in computational biology, where one aims at computing a lowrank matrix factorization involving a binary factor. In a random model with adjustable expected sparsity of the binary factor, we show for a large class of random binary factors that the corresponding factorization problem is uniquely solvable with high probability. In data analysis, such uniqueness results are of particular interest; ambiguous solutions often lack interpretability and do not give an insight into the structure of the underlying data. For proving uniqueness in this random model, small ball probability estimates are a key ingredient. Since to the best of our knowledge, there are no such estimate suitable for our application, we prove an extension of the famous Lemma of Littlewood and Offord. Hereby, we also discover a connection between the matrix factorization problem at hand and the notion of Sperner families. 
Room: 
02.10.011 
Tuesday, 29.11.2016 
16:0017:00 
Speaker: 
Christian Kümmerle, Juliane Sigl (TUM) 
Title: 
We propose a new Iteratively Reweighted Least Squares (IRLS) algorithm for the problem of recovering a lowrank matrix from incomplete linear observations. Unlike in previous IRLS approaches for the problem, we exploit the information in the column as well as in the row space of the iterates, using socalled harmonic mean weight matrices. Our nonconvex algorithm exhibits local convergence guarantees comparable to previous IRLS algorithms for generic random measurement operators with a number of measurements in the optimal order. Additionally, we prove that the exhibited rate of convergence is superlinear, which is unprecedented even for other iterative approaches for the lowrank recovery problem. Our theoretical findings align very well with our numerical experiments, which also suggest that convergence to nonglobal minimizers is not an issue and that the proposed algorithm needs fewer measurements for successful recovery than any other algorithm proposed in the literature. 
Room: 
02.12.020 
Tuesday, 22.11.2016 
16:0017:00 
Speaker: 
YoungPil Choi ^{} (TUM) 
Title: 
The jargon "flocking" denotes the phenomenon in which selfpropelled individuals using only limited environmental information and simple rules organize into an ordered motion. In this talk, we study the CuckerSmale flocking model which is a type of Newtonian equations. We first present the largetime behavior of solutions and discuss modified CuckerSmale models that address several drawbacks of the original model. In the second part, we deal with the rigorous derivation of kinetic CuckerSmale type equations when sharp sensitivity regions are considered. 
Room: 
02.12.020 
Tuesday, 15.11.2016 
Speaker: 
Kristof Schröder ^{} (Helmholtz Zentrum München) 
Title: 
We will look at the possibility of recovering a sum of Dirac measures on the rotation group SO(3) from its low degree moments with respect to WignerD functions only. Topics in the talk will cover theoretical results for recovery guarantee and algorithmic aspects of the problem. 
Room: 
02.10.011 
Tuesday, 08.11.2016 
Speaker: 
Jakob Geppert (Göttingen) 
Title: 
Improved convergence guarantees for Sparse Power Factorization 
Room: 
02.10.011 
Tuesday, 25.10.2016 
Speaker: 
Matthias Wissel (TUM) 
Title: 
We will propose a timedependent nonhomogeneous Markov model for predicting free parking spaces in urban areas. We therefore will describe the formalized general setting that reflects the real world environment we work in when concerned with parking prediction. Also, issues concerning the matter of availability of the right kind of data will be addressed. Then, the parameter estimation based on a given data set will be of particular interest. To this end, an extended framework of empirical risk minimization for matrixvalued functions and the use of matrixvalued reproducing kernel Hilbert spaces to approximate the timedependent generator of the Markov process will be introduced. Finally, we want to give an outlook on further research possibilities in this field with a focus on a purely data driven approach. 
Room: 
02.10.011 
Tuesday, 18.10.2016 
Speaker: 
Michael Sandbichler (TUM) 
Title: 
Efficient data representation is a core aspect of modern signal processing, for example sparsity of data has been very efficiently used in compressed sensing in order to improve the required minimal number of measurements. We will look at a 'dual' version of sparsity, namely Analysis cosparsity and provide new sequential algorithms to find Analysis Operators which sparsify a given class of signals. 
Room: 
02.10.011 
Contact
top
