Kolloquientermine

Kolloq. Dr. Todor Mladenov, topic: Cellular Technology Evolution: The path to 5G

Wednesday, 13th of November 2019, 14:15 pm, 01.07.023 (FMI, Campus Garching)

Abstract:

The talk will present the evolution of modem cellular technology from the early days of 1G connectivity to today’s 5G systems. Special emphasis would be given to the trade-off in decision making and the influence of politics and economy on technology. The talk will further provide a detailed overview of the 4G and 5G physical layer downlink and uplink signal processing and the architectural challenges related to it.

Bio:

Dr. Todor Mladenov is a HW baseband modem architect at Intel Corporation responsible for the architecture and design of the 4G and 5G physical layer modem hardware. Since joining Intel in 2012, Dr. Mladenov defined and designed the ARM and x86 processor subsystems used in the XMM7x60 and IBIS products and helped create MNoC, the first PCI ordered and AMBA compliant scalable network on chip fabric. For this work in 2015 he received an Intel Division Recognition Award. Before joining Intel, Dr. Mladenov was an IT professor with the Communications Department of Gwangju Institute of Science and Technology (GIST) where he thought courses and conducted research in information theory, brain computer interface (BCI) and embedded systems. He has published eight journal papers, eighteen conference papers and has three filed patents. Dr. Mladenov received his BE from Technical University of Sofia (TU-Sofia) and his MS and Ph.D. degrees from GIST.

Contact person:

Linus Dietz, M.Sc.

Phone: +49.89.289.18676

Email: linus.dietz@tum.de

 

 

 

 

Kolloq. Serge Egelman, topic: Empowering Users to Make Privacy Decisions in Mobile Environments

Wednesday, 7th of August 2019, 16:00 pm, HS 3 (FMI, Campus Garching)

Abstract:

Mobile platforms have enabled third-party app ecosystems that provide users with an endless supply of rich content. At the same time, mobile devices present very serious privacy risks: their ability to capture real time data about our behaviors and preferences has created a marketplace for user data that most consumers are simply unaware of. In this talk, I will present prior and ongoing research that my group has performed to understand how users make privacy decisions on their mobile devices, including work that we have done to improve the usability of the permission-granting process through the use of machine learning. I will also present research that my research group has conducted to automatically examine the privacy behaviors of mobile apps. Using analysis tools that we developed, we have tested over 100,000 of the most popular Android apps to examine what data they access and with whom they share it. I will present data on how mobile apps are tracking and profiling users, how these practices are often against users' expectations and public disclosures, and how app developers may be violating various privacy regulations.

Bio:

Serge Egelman is the Research Director of the Usable Security and Privacy group at the International Computer Science Institute (ICSI), which is an independent research institute affiliated with the University of California, Berkeley. He conducts research to help people make more informed online privacy and security decisions, and is generally interested in consumer protection. This has included improvements to web browser security warnings, authentication on social networking websites, and most recently, privacy on mobile devices. Seven of his research publications have received awards at the ACM CHI conference, which is the top venue for human-computer interaction research; his research on privacy on mobile platforms has been cited in numerous lawsuits and regulatory actions, as well as featured in the New York Times, Washington Post, Wall Street Journal, Wired, CNET, NBC, and CBS. He received his PhD from Carnegie Mellon University and has previously performed research at Xerox Parc, Microsoft, and NIST.

 

Contact person:

Prof. Jens Grossklags, Ph.D.

Phone: +49.89.289.17745

Email: jens.grossklags@tum.de

 

 

 

 

Kolloq. Prof. Florian Kerschbaum, topic: Secure Analytics

Tuesday, 30th of July 2019, 15:00 pm, Fakultätsraum Informatik 00.12.019 (FMI, Campus Garching)

Abstract:

This talk covers two recent works in computing analytical functions over encrypted data. I will start with the traditional range search problem and show how to encrypt data, such that it can be efficiently searched and stored, but is provably secure against attacksby a snap-shot attackers. The problem entails a definition of securely encrypting a datastructure. Our approach can withstand all recent attacks on property-preservingly encrypted databases. In the second part of the talk I will talk about the problem of computing rank functions over the blockchain. This can be used to implement sealed-bid auctions over the blockchain. Our approach is secure against malicious adversaries and requires only 3 to 4 blocks to compute. We combine a numberof techniques in an usual way to achieve this trade-off.

Bio:

Florian is an associate professor in the David R. Cheriton School of Computer Science at the University of Waterloo (since 2017) and executive director of the Waterloo Cybersecurity and Privacy Institute (since 2018). Before he worked as chief research expert at SAP in Karlsruhe (2005 – 2016) and as a software architect at Arxan Technologies in San Francisco (2002 – 2004). He holds a Ph.D. in computer science from the Karlsruhe Institute of Technology (2010) and a master's degree from Purdue University (2001). He isinterested in data security and privacy in data management, machine learning, and blockchains. He extends real-world systems with cryptographic security mechanisms to achieve (some) provable security guarantees.

 

Contact person:

Prof. Georg Carle

Phone: +49.89.289.18030

Email: carle@in.tum.de

 

 

 

 

 

Special Kolloq. Christian Freksa, topic: Spatial Problem Solving - from Konrad Zuse's 'Rechnender Raum' to 'Strong Spatial Cognition'

Friday, 19 th of July 2019, 16:00 pm, FMI Room 00.13.009A (MI-Building, Campus Garching)

Abstract:

Fifty years ago Konrad Zuse published the book Rechnender Raum ("Calculating Space").He proposed a discrete physics of space that would permit the use of structure-preserving precise representations in place of discrete approximation of continuous physics. I will briefly review Zuse's proposal before presenting the Strong Spatial Cognition (SSC) paradigm to geometric problem solving. SSC pursues the preservation of structures by avoiding the concept mapping problem in a different way. The paradigm involves perceiving and acting agents (such as humans and robots) that make direct use of spatial structures and spatial affordances in their environment (knowledge in the world) instead of operating on a digital twin of the world. In the approach, computation shifts from object-level reasoning by means of spatial calculi to meta-level reasoning that controls perception and operations in space. I will present spatial problems and how they are solved in SSC and explain how the cognitively motivated approach avoids the computational complexity trap in spatial reasoning. I will argue that the paradigm is suitable for future autonomous robot systems and may have uses well beyond the spatial domain.

Bio:

Christian Freksa is a Research Professor of Cognitive Systems at the Faculty of Mathematics and Informatics, University of Bremen. His research concerns representationand reasoning with incomplete, imprecise, lean, coarse, approximate, fuzzy, and conflicting knowledge about spatial environments. Freksa received a PhD in Artificial Intelligence from UC Berkeley. Before joining the University of Bremen he carried out research at the Max Planck Institute for Psychiatry and the Technical University of Munich, the International Computer Science Institute in Berkeley, and the University of Hamburg. From 1996 to 2014 he directed the DFG priority program 'Raumkognition' and the SFB/Transregio 'Spatial Cognition'.

 

contact person:

Helma Piller

Phone: +49.89.17300

Email: piller@in.tum.de

 

 

 

 

Kolloq. Marc Brockschmidt, topic: Learning from Programs as Graphs

Wednesday, 17th of July 2019, 2:15 pm, Interims HS 1 (Campus Garching)

Abstract:

Deep Learning has been the crucial step forward in perceptual tasks such as the understanding of images, speech and natural language. So far, it has had less impact on the practice of Software Engineering, where it is competing with a wide variety of mature, existing methods based on logic and deduction. I will discuss how these two worlds can be combined using graph-based representations, with applications to finding bugs and generating new source code.

Bio:

Marc Brockschmidt is a Senior Researcher in the All Data AI and Programming Principles and Tools groups at Microsoft Research Cambridge (UK). He obtained his PhD from RWTH Aachen University studying formal methods that can automatically prove termination of programs. Surprisingly, that worked substantially less well than manually proving termination of programs. He thus moved on to study how computers can learn the skills that make humans better at programming than machines.

 

Contact person:

Prof. Stephan Günnemann

Phone: +49.89.289.17282

Email: guennemann@in.tum.de

 

 

 

 

Kolloq. Prof. Vijay Vazirani, topic: Matching is as Easy as the Decision Problem, in the NC Model

Wednesday, 17 th of July 2019, 11:15 am, FMI Room 01.10.011 (MI-Building, Campus Garching)

Abstract:

Is matching in NC, i.e., is there a deterministic fast parallel algorithm for it? This has been an outstanding open question in TCS for over three decades, ever since the discovery of Random NC matching algorithms. Over the last five years, the TCS community has launched a relentless attack on this question, leading to the discovery of numerous powerful ideas. We give what appears to be the culmination of this line of work: An NC algorithm for finding a minimum weight perfect matching in a general graph with polynomially bounded edge weights, provided it is given an oracle for the decision problem. Consequently, for settling the main open problem, it suffices to obtain an NC algorithm for the decision problem. We believe this new fact has qualitatively changed the nature of this open problem. All known efficient matching algorithms for general graphs follow one of two approaches: those initiated by Edmonds (1965) or Lovasz (1979). Our algorithm is based on a new approach which was inspired by the multitude of ideas discovered in the last five years. The difficulty of obtaining an NC perfect matching algorithm led researchers to study matching vis-a-vis clever relaxations of the class NC. In this vein, Goldwasser and Grossman (2015) gave a pseudo-deterministic RNC algorithm for finding a perfect matching in a bipartite graph, i.e., an RNC algorithm with the additional requirement that on the same graph, it should return the same (i.e., unique) perfect matching for almost all choices of random bits. A corollary of our reduction is an analogous algorithm for general graphs.

This talk is fully self-contained.

Based on the following joint paper with Nima Anari:

arxiv.org/pdf/1901.10387.pdf

 

Bio:

Vijay Vazirani is a Distinguished Professor at UC Irvine.

en.wikipedia.org/wiki/Vijay_Vazirani

www.ics.uci.edu/~vazirani/

 

contact person:

Susanne Albers

Phone: +49.89.289.17702

Email: albers@in.tum.de

 

 

 

 

Kolloq. Prof. Sven Seuken, topic: “Machine Learning-powered Iterative Combinatorial Auctions"

8th July 2019, 14:30 pm, IAS Faculty Club (left),  Garching

Abstract:

In this talk, I will discuss how machine learning can be used to improve the design of market mechanisms. I will present one particular case study in detail: a machine learning-powered iterative combinatorial auction (CA). The main goal of integrating machine learning (ML) into the auction is to improve preference elicitation, which is a major challenge in large CAs. In contrast to prior work, our auction design uses "value queries" instead of prices to drive the auction. The ML algorithm is used to help the auction mechanism decide which value queries to ask in every iteration. While using ML inside an auction introduces new challenges, we demonstrate how we obtain an auction design that is individually rational, has good incentives, and is computationally tractable. Via simulations, we benchmark our new auction against the well-known combinatorial clock auction (CCA). Our results demonstrate that the ML-powered auction achieves higher allocative efficiency than the CCA, even with only a small number of value queries. *Joint work with Gianluca Brero (University of Zurich) and Benjamin Lubin (Boston University).

Short Bio:

Sven Seuken is a tenured Associate Professor at the Department of Informatics of the University of Zurich and head of the Computation and Economics Research Group. His research lies at the intersection of Computer Science and Game Theory, with a focus on market design. He received his Ph.D. in Computer Science from Harvard University in 2011 under the guidance of David C. Parkes. Since joining the University of Zurich, he has received several awards and research grants, including a Google Faculty Research Award and an ERC Starting Grant. In 2017, he was ranked as one of the "Top 40 under 40" in the category "society and science" by the German business magazine Capital. He is also interested in practical applications of market design: he is the Chief Economist at BandwidthX, Inc. (USA) and he is a senior market design advisor for Tremor Technologies, Inc. (USA).

contact person:

Martin Bichler

Phone: +49.89.289.17500

Email: bichler@in.tum.de

 

 

 

 

Kolloq. Benjamin Seibold, topic: Instabilities in Homogeneous and Heterogeneous Traffic Flow

Friday, 5th of July 2019, 11:00 am, FMI 01.09.014 (MI Building, Campus Garching)

Abstract:

A fundamental property of vehicular traffic flow is that it mayexhibit significant non-equilibrium behavior once the flow becomes sufficiently congested: vehicles do not move at steady speeds, but rather undergo acceleration and braking. This unsteadiness manifests in a multitude of ways; but in almost all situations it is undesirable (loss of efficiency; increased fuel consumption, emissions, accident risk). Understanding the causes and dynamics of non-equilibrium traffic flow via models, observations, and data is a crucial challenge in traffic flow theory; and it becomes even more important in light of ongoing and future (partial) automation of the vehicle fleet. We highlight important steps in developing phenomenological models for instabilities and nonlinear waves in traffic flow and discuss how this modeling extends to heterogeneous flows composed of a mix of human-driven and (semi-)automated vehicles. In particular, we stress how automation can go both ways: it could make traffic run more smoothly; but also render it less efficient than it is under human control.

Bio:

Dr. Benjamin Seibold is an Associate Professor in the Department ofMathematics at Temple University (Philadelphia), and the Director of the Center of Computational Mathematics and Modeling. He received his PhD from the University of Kaiserslautern, and was an Instructor at MIT. He works in applied and computational mathematics with specific applications in traffic flow, fluid dynamics, radiation transport, and life science applications. His traffic flow research focuses on the understanding of non-equilibrium phenomena (such as stop-and-go waves), particularly in mixed human-automated traffic flow, via modeling, simulation, and experiments.

contact person:

Raphael Stern

Email: raphaelestern@gmail.com

 

 

 

 

Kolloq. Prof. Yue Zhang, topic: Interactive Visualization of Planar Kaleidoscopic Orbifolds

Friday, 5 th of July 2019, 11:00 am, FMI Room 02.09.023 (MI-Building, Campus Garching)

Abstract:

Orbifold is a modern mathematical concept that has been used to understand the geometric structures of hyperbolic geometry and prove the famous Poincar\'e conjecture for the three-dimensional case (the last and the hardest case). Orbifolds contain intricate structures which not only render orbifolds an interesting subject but also make their understanding challenging. In this paper, we provide an interactive visualization system for a class of important orbifolds: {\em kaleidoscopic orbifolds}. With the system, the user can create kaleidoscopic scenes of arbitrary complexity and interact with the objects in the scene to gain critical insights on kaleidoscopic orbifolds. Our visualization techniques are based on mirror reflections, a metaphor that is conceptually well understood by an average user. Furthermore, we develop interactive games to help the user better understand the properties of kaleidoscopic orbifolds. Our visualization system and interaction techniques are useful to gain intuitive comprehension of important concepts and properties related to orbifolds such as groups, group actions, branched covering spaces, and the result that all planar kaleidoscopes have a zero Euler characteristic. To test the efficiency of our system, we have conducted a user study, with the users being high school and college students as well as professors in mathematics teaching differential geometry, abstract algebra, and topology.

Bio:

http://eecs.oregonstate.edu/computer-graphics-and-visualization

 

contact person:

Nils Thürey

Phone: +49.89.289.19484

Email: nils.thuerey@tum.de

 

 

 

 

Kolloq. Prof. Eugene Zhang, topic: Tensor Field Design in Volumes

Thursday, 4 th of July 2019, 15:00 pm, FMI Room 02.13.010 (MI-Building, Campus Garching)

Abstract:

3D tensor field design is important in several graphics applications such as procedural noise, solid texturing, and geometry synthesis. Different fields can lead to different visual effects. The topology of a tensor field, such as degenerate tensors, can cause artifacts in these applications. Existing 2D tensor field design systems cannot be used to handle the topology of a 3D tensor field. In this paper, we present to our knowledge the first 3D tensor field design system. At the core of our system is the ability to edit the topology of tensor fields. We demonstrate the power of our design system with applications in solid texturing and geometry synthesis.

 

Bio:

https://web.engr.oregonstate.edu/~zhange/

 

contact person:

Nils Thürey

Phone: +49.89.289.19484

Email: nils.thuerey@tum.de

 

 

 

 

 

Kolloq. Evangelos Pournaras, topic: Decentralized Collective Learning for Self-managed Sharing Economies

Thursday, 6 th of June 2019, 10:00 am, FMI Room 01.11.018 (MI-Building, Campus Garching)

Abstract:

The Internet of Things equips citizens with phenomenal new means for online participation in sharing economies. When agents self-determine options from which they choose, for instance their resource consumption and production, while these choices have a collective system-wide impact, optimal decision-making turns into a combinatorial optimization problem known as NP-hard. In such challenging computational problems, centrally managed (deep) learning systems often require personal data with implications on privacy and citizens’ autonomy. This work envisions an alternative unsupervised and decentralized collective learning approach that preserves privacy, autonomy and participation of multi-agent systems self-organized into a hierarchical tree structure. Remote interactions orchestrate a highly efficient process for decentralized collective learning. This disruptive concept is realized by I-EPOS, the Iterative Economic Planning and Optimized Selections, accompanied by a paradigmatic software artifact. Strikingly, I-EPOS outperforms related algorithms that involve non-local brute-force operations or exchange full information. This talk illustrates experimental findings about the influence of network topology and planning on learning efficiency as well as findings on techno-socio-economic trade-offs and global optimality. Experimental evaluation with real-world data from energy and bike sharing pilots demonstrates the grand potential of collective learning to design ethically and socially responsible participatory sharing economies.

Bio:

Dr. Evangelos Pournaras is a senior scientist in the Professorship of Computational Social Science, at ETH Zurich, Zurich, Switzerland. He was earlier at VU University Amsterdam and Delft University of Technology in the Netherlands, where he completed his PhD studies in 2013. Since 2007, he holds a MSc with distinction in Internet Computing from University of Surrey, UK and since 2006 a BSc on Technology Education and Digital Systems from University of Piraeus, Greece. Evangelos has also been a visiting researcher at EPFL in Switzerland and has industry experience at IBM T.J. Watson Research Center in the USA. Currently he is also involved with the enterprise blockchain platform of Insolar as a research fellow. He serves the editorial board and the program committees of several international conferences and journals. He has several publications in high-impact journals and conferences, including a best journal paper award. He has managed and contributed to several EU funded projects such as ASSET, SoBigData and FuturICT 2.0. His research focuses on decentralized self-managed systems applied in techno-socio-economic application domains of Smart Grids and Smart Cities.

contact person:

Ilias Gerostathopoulos

Phone: +49.89.289.17378

Email: gerostat@in.tum.de

 

 

 

 

Kolloq. Prof. Benny Moldovanu, topic: "A Theory of Auctions with Endogenous Valuations"

Monday, 3 rd of June 2019, 14:00 pm, IAS  Ground Floor, "Audi" (Lichtenbergstr. 2a, Campus Garching)

Abstract:

We derive the symmetric, revenue maximizing allocation of m units among n symmetric agents who have unit demand, and who take costly actions that influence their values before participating in the mechanism. The auction with costly actions can be represented by a reduced form model where agents have convex, non-expected utility preferences over the interim probability of receiving an object. Both the uniform, m+1-price auction and the discriminatory pay-your-bid auction with reserve prices constitute symmetric revenue maximizing mechanisms. Contrasting the case with exogenous valuations, the optimal reserve price reacts to both demand and supply, i.e., it depends both on the number of objects m and on number of agents n. The main tool in our analysis is an integral inequality involving majorization, super-modularity and convexity due to Fan and Lorentz .

(joint work with Alex Gershkov, Hebrew U. Jerusalem and Philipp Strack, UC Berkeley)

Bio:

Benny Moldovanu ist Professor für Wirtschaftswissenschaften an der Universität Bonn, Direktor der dort ansässigen Graduiertenschule und Gründungsmitglied sowie Co-Direktor des Hausdorff-Zentrums für Mathematik und des Reinhard Selten Institute. Als Gastprofessor war er zudem an den Universitäten Michigan, Northwestern, Yale, Tel-Aviv, Jerusalem sowie am University College London tätig. Er ist Fellow der Econometric Society, der European Economic Association und der Game Theory Society. Moldovanu erhielt den Max-Planck-Preis sowie den Gossen-Preis. Er war Mitherausgeber von Econometrica, des Journal of Economic Theory, Games and Economic Behavior und des Journal of the European Economic Association. Seine Forschung konzentriert sich auf das Mechanism Design und dessen Anwendung auf Auktionen, dynamische Preisgestaltung, Wettbewerbe sowie Wahlverfahren. Darüber hinaus hat er große Unternehmen und Regierungen beim Design von Auktionen sowie in der Strategieplanung beraten.

 

ontact person: Martin Bichler

Phone: +49.89.289.17500

Email: bichler@in.tum.de

 

 

 

Kolloq. Philip Paré, topic: Virus Spread over Networks

Monday, 3 rd of June 2019, 14:00 pm, FMI Room 01.11.018 (MI-Building, Campus Garching)

Abstract:

The study of epidemic processes has been a topic of interest for many years over a wide range of areas, including computer science, mathematical systems, biology, physics, social sciences, and economics. More recently, there has been a resurgence of interest in the study of epidemic processes focused on the spread of viruses over networks, motivated not only by security threats posed by computer viruses, but also recent devastating outbreaks of infectious diseases and the rapid spread of opinions over social networks. Up to this point these network-dependent spread models have not been validated by real data. In this talk, we analyze a mathematical model for network-dependent spread and use that analysis to identify the healing and infection parameters of the model. We apply these ideas, employing John Snow's seminal work on cholera epidemics in London in the 1850's, to validate the susceptible-infected-susceptible (SIS) model. The validation results are surprisingly good, capturing the behavior of the cholera epidemic from John Snow's 1854 dataset quite well. We conclude by briefly highlighting extensive analysis and algorithm design results we have obtained for time-varying and multi-layered networks, and finally discuss various directions for compelling future work.

Bio:

Philip E. Paré received his B.S. in mathematics with University Honors and his M.S. in Computer Science from Brigham Young University, Provo, UT, in 2012 and 2014, respectively, and his Ph.D. in Electrical and Computer Engineering (ECE) from the University of Illinois at Urbana-Champaign (UIUC), Urbana, IL in 2018. He is currently a postdoctoral scholar in the Division of Decision and Control Systems in the School of Electrical Engineering and Computer Science at KTH Royal Institute of Technology in Stockholm, Sweden. Philip was the recipient of the 2017-2018 Robert T. Chien Memorial Award for excellence in research from the UIUC ECE Department and named a 2017-2018 UIUC College of Engineering Mavis Future Faculty Fellow. His research interests include the modeling, control, and security of dynamic networked systems, biological systems, and time--varying systems.

contact person:

Raphael Stern

Phone: +49.89.289.17800

Email: raphael.stern@tum.de

 

 

 

 

Kolloq. Dr. Sunghan Ryu, topic: The Effect of Crowdfunding Success on Subsequent Financing Outcomes of Start-ups

Wednesday 15th of May 2019, 3:00 pm, FMI 01.13.010 (MI-Building, Campus Garching)

Abstract:

From the perspective of signaling theory, this study examines how the receipt of crowdfunding (compared with angel investing) is associated with start-ups’ subsequent financing outcomes. We collected data on crowdfunded start-ups as well as angel-funded start-ups and their subsequent financing from venture capitalists. Our results, after addressing the potential endogeneity using a bivariate probit model and propensity score matching, show that crowdfunded start-ups and angel investing start-ups have no statistically significant difference in receiving subsequent venture capital (VC) investments. Interestingly, however, the effect of obtaining crowdfunding on the receipt of subsequent investments from VCs differs across different characteristics of startup-ups. Moreover, when we compare corporate venture capitalists (CVCs) with independent venture capitalists (IVCs), obtaining crowdfunding is positively associated with the receipt of subsequent investments from CVCs, but not from IVCs.

Bio:

Dr. Sunghan Ryu is an assistant professor of USC-SJTU Institute of Cultural and Creative Industry at Shanghai Jiao Tong University. He earned his Ph.D. in IT management from College of Business, KAIST. His research interests include IT innovations in cultural and creative industries and effective information systems practices in the entrepreneurial context. His current research interests revolve around user behaviors in crowdfunding platforms and other various multi-sided platforms. His work appeared in academic journals including Journal of Strategic Information Systems, Electronic Markets, Electronic Commerce Research & Applications and presented at several prestigious conferences such as ICIS, PACIS, and HICSS.

contact person:

Dr. Manuel Wiesche

Phone: +49.89.289.19539

Email: manuel.wiesche@in.tum.de

 

 

 

 

Kolloq. Tim Kraska, topic: Towards Learned Algorithms, Data Structures, and Systems

Monday, 13 th of May 2019, 14:00 pm, FMI Room 02.09.014 (MI-Building, Campus Garching)

Abstract:

All systems and applications are composed from basic data structures and algorithms, such as index structures, priority queues, and sorting algorithms. Most of these primitives have been around since the early beginnings of computer science (CS) and form the basis of every CS intro lecture. Yet, we might soon face an inflection point: recent results show that machine learning has the potential to alter the way those primitives are implemented and the performance they can provide for specific applications. In this talk, Tim Kraska will outline different ways to build learned algorithms and data structuresto achieve “instance-optimality” with a particular focus on techniques used as part of data management systems.

Bio:

Tim Kraska is an Associate Professor of Electrical Engineering and Computer Science in MIT's Computer Science and Artificial Intelligence Laboratory and co-director of the Data System and AI Lab at MIT (DSAIL@CSAIL). Currently, his research focuses on building systems for machine learning, and using machine learning for systems. Before joining MIT, Tim was an Assistant Professor at Brown, spent time at Google Brain, and was a PostDoc in the AMPLab at UC Berkeley after he got his PhD from ETH Zurich. Tim is a 2017 Alfred P. Sloan Research Fellow in computer science and received several awards including the 2018 VLDB Early Career Research Contribution Award, the 2017 VMware Systems Research Award , an NSF CAREER Award, as well as several best paper and demo awards at VLDB and ICDE.

contact person:

 André Kohn

Email: andre.kohn@tum.de

 

 

 

 

Kolloq. Prof. Makoto Kaneko , topic: Beyond Human Technology” Opens a New Bio World

Friday, 22 nd of February 2019, 14:00 pm, FMI Fakultätsraum Informatik Room 00.12.019 (MI-Building, Campus Garching)

Abstract:

This talk begins by explaining what is“Beyond Human”. Knowing of the limitation of human perception and action, we show how to design an artificial system leading to “Beyond Human” by utilizing two kernel components, an online high speed vision and a high speed actuator where both speeds are several hundred times faster than human eye and muscle, respectively. We show a couple of examples of “Beyond Human Robot”. As for bio application, we show fast and fine cell manipulation system with the frequency of 100Hz and the resolution of 250 nanometers by using both a newly developed syringe pump and an online high speed vision. As an application of cell manipulation, we show “Cell Stress Test” where a mechanical stress is continuously imparted to a cell until it eventually gets damages. We also show an interesting behavior of red blood cells where their recovery characteristics after three-minutes-loading in microchannel dramatically change. Finally, we show our hypothesis where our brain activity correlates with the deformability of red blood cell. All topics in this talk will be explained together with video demonstration.

Bio:

Dr. Makoto Kaneko (M’88-SM’03-F’06 for IEEE) is a professor of the Department of Mechanical Engineering, Graduate School of Engineering, Osaka University. He received Ph.D. at the University of Tokyo in 1981. His current research interests include dynamic active sensing, such as Strobe Imager, cell deformability sensing, dynamic sensing of human eye, and dynamic sensing of internal organs by using both high speed vision and high speed actuator. He has received 30 awards, including the Humboldt Research Award in 1997, the IEEE ICRA Best Manipulation Paper Award in 2000, the IEEE ISATP Outstanding Paper Award in 2001, the IEEE RAS King-Sun Fu Memorial Best Transactions Paper Award in 2003, the IEEE ICIA Best Conference Paper Award in 2005, the IEEE ICMA Best Paper Award in Automation in 2013, the IEEE MHS Best Paper Award in 2012 and 2014, IEEE Int. Conf. on Mechatronics and Automation, the Toshio Fukuda Award in 2015, and IEEE Int. Conf. on Cyborg and Bionic Systems, the Best Paper Award in 2017. He also received the Honorary Doctor from Darmstadt University of Technology, Germany in 2013.

contact person:

Prof. Alois KNOLL / Amy Bücherl

Phone: +49.89.289.18110

Email: buecherl@tum.de

 

 

 

 

Kolloq. Leonardo Linguaglossa, topic: High-speed NFV: performance evaluation and modeling

Monday, 18 th of February 2019, 17:00 pm, FMI Room 03.07.023 (MI-Building, Campus Garching)

Abstract:

Network Functions Virtualization (NFV) is among the latest network revolutions, bringing flexibility and avoiding network ossification. While NFV provides a flexible way of implementing network functions on commodity hardware, an all-software NFV implementation may present a performance gap with respect to hardware-based solutions. In the last decade numerous software acceleration techniques have appeared to bring high-speed capabilities to software network frameworks, thus trying to reduce the distance w.r.t. pure hardware solutions. Batching is one example of such techniques, consisting in processing packets in groups as opposed to individually, which is required at high-speed to minimize the framework overhead, reduce interrupt pressure, and leverage instruction-level cache hits. Whereas several system implementations have been proposed and experimentally benchmarked, the scientific community has so far only to a limited extent attempted to model the system dynamics of modern NFV routers exploiting batching acceleration. We fill this gap by proposing a simple generic model for such batching-based mechanisms, which allows a very detailed prediction of highly relevant performance indicators. These include the distribution of the processed batch size as well as queue size, which can be used to identify loss-less operational regimes or quantify the packet loss probability in high-load scenarios. In this talk I will present our experimental campaign for performance evaluation using a state-of-the-art NFV router, namely VPP. Then I will introduce our model for a generic NFV router. We contrast the model prediction with experimental results gathered in our testbed, showing that the model not only correctly captures system performance under simple conditions, but also in more realistic scenarios in which traffic is processed by a mixture of functions.

Bio:

Leonardo Linguaglossa is currently a post-doctoral researcher at Telecom PairsTech (France) working in a collaboration with Cisco named "NewNet@Paris". In 2018/2019 he is working in a joint collaboration between TPT and TUM with a project named "AI4P" (Artificial Intelligence for Performance). Leonardo's main research interests include high-speed networking, future network architectures (NFV, SDN), performance evaluation and modeling.

contact person:

Prof. Georg Carle / Veronika Fleischner

Phone: +49.89.289.18032

Email: fleischner@net.in.tum.de

 

 

 

 

Kolloq. Prof. Christine Legner , topic: Managing Data as an Asset - Creating the Foundations of the Digital and Data-Driven Enterprise

Friday 15th of February 2019, 1:00 pm, FMI 01.13.010 (MI-Building, Campus Garching)

Abstract:

In the digital and data-driven economy, data is evolving into a strategic resource for enterprises in all industries. However, data’s growing role and increasing business criticality is not yet reflected in today’s management practice. Many enterprises face challenges relating to poor data quality, the existence of data silos in organizations, and increasing regulatory burdens. We propose a reference model that supports digital and data-driven enterprises in managing data as a strategic resource. The reference model accumulates academic and practical knowledge in the data management field. It has been developed in a unique industry-research collaboration involving more than 30 European companies and researchers from three universities over a period of 12 years. Based on the understanding of data as a strategic resource, the reference model conceptualizes data management as goal and outcome-oriented capabilities that contribute to business capabilities and are developed in a continuous improvement cycle. The development of the reference model for data management in this longitudinal research setting provides opportunities to reflect on the accumulation and evolution of knowledge. We find that knowledge accumulation and evolution occur as result of maturing design knowledge and evolving practical challenges, and that our artifact development materialized in the stages from invention to refinement, and transfer/exaptation to new problem spaces.

Bio:

Christine Legner is Full Professor of Information Systems at HEC Lausanne, University of Lausanne, where she teaches business information systems and enterprise architecture. She is also the academic director of the Competence Center Corporate Data Quality (CC CDQ), a research consortium and expert community in the field of corporate data management. Her research interests revolve around IT-enabled business innovations, resulting from the convergence of cloud, mobile and analytical technologies. She works on concepts and methods for corporate data management and strategic IT planning to align business information systems with organizational goals and structures. Before joining HEC Lausanne, Christine Legner was professor at European Business School in Wiesbaden (Germany). She holds a PhD and post-doctoral qualification (“Habilitation”) in Information Systems from the University of St. Gallen (Switzerland). She has been visiting scholar at INSEAD, Stanford University and University of Montreal.

contact person:

Dr. Manuel Wiesche

Phone: +49.89.289.19539

Email: manuel.wiesche@in.tum.de

 

 

 

 

Kolloq. Rens van der Heijden , topic: Misbehavior Detection in Cooperative Intelligent Transport Systems and Beyond

Wednesday 6th of February 2019, 10:00 am, FMI 01.11.018 (MI-Building, Campus Garching)

Abstract:

In recent years, significant progress has been made in the domains of autonomous driving and vehicular inter-connectivity. Many envision a future in which transportation is partially or even fully automated; to achieve such a vision, autonomous vehicles need a means of cooperation and information sharing. Cooperative Intelligent Transport Systems (C-ITS) are a major step in this direction, where the needs of autonomous driving are explicitly considered in the networking community. This leads to applications such as cooperative adaptive cruise control (CACC), in which vehicles share information to create a platoon with minimal safety distance between the vehicles. Although security has received some attention, it has largely been limited to message integrity, authenticity, and privacy considerations.This talk will give an introduction to the security of C-ITS, with a detailed overview of misbehavior detection, which refers to the detection of attacks that directly affect application behavior, such as false data injection. These attacks are of particular interest, because they cannot be prevented by traditional cryptographic approaches, since in this attacker model, the attacker possesses valid key material. I will introduce Maat, a framework for misbehavior detection and fusion that aims to detect such attacks. A brief outlook will discuss future challenges, such as the integration of this detection framework with sensor data fusion and attack mitigation strategies.

Bio:

Dr. Rens W. van der Heijden defended his PhD thesis on November 9th, 2018 at Ulm University, working on Misbehavior Detection in Cooperative Intelligent Transport Systems at the Institute of Distributed Systems. He continues to work at the institute as a PostDoc, working on misbehavior detection, security, and various automotive topics. Rens has previously received a Bachelor diploma in Computer Science from the Universty of Twente in 2010 and a Master diploma cum laude from the Kerckhoffs Institute at the Dutch Universities Twente, Nijmegen, and Eindhoven in 2012.

contact person:

Severin Kacianka

Phone: +49.89.289.17340

Email: kacianka@in.tum.de

 

 

 

 

Kolloq. Oriol Vinyals, topic: AlphaStar: Mastering the Real-Time Strategy Game StarCraft II

Thursday 31st of January 2019, 06:00 pm, FMI 5602.EG.001HS 1, Friedrich L. Bauer Hörsaal (MI-Building, Campus Garching)

Abstract:

Games have been used for decades as an important way to test and evaluate the performance of artificial intelligence systems. As capabilities have increased, the research community has sought games with increasing complexity that capture different elements of intelligence required to solve scientific and real-world problems. In recent years, StarCraft, considered to be one of the most challenging Real-Time Strategy (RTS) games and one of the longest-played esports of all time, has emerged by consensus as a “grand challenge” for AI research.

Bio:

Oriol Vinyals is a Research Scientist at Google DeepMind, working in Deep Learning. Prior to joining DeepMind, Oriol was part of the Google Brain team. He holds a Ph.D. in EECS from the University of California, Berkeley and is a recipient of the 2016 MIT TR35 innovator award. His research has been featured multiple times at the New York Times, BBC, etc., and his articles have been cited over 29000 times. His academic involvement includes program chair for the International Conference on Learning Representations (ICLR) of 2017, and 2018. Some of his contributions are used in Google Translate, Text-To-Speech, and Speech recognition, used by billions. At DeepMind he continues working on his areas of interest, which include artificial intelligence, with particular emphasis on machine learning, deep learning and reinforcement learning.

 

 

contact person:

Prof. Dr. Matthias Niessner

Phone: +49.89.289.19556

Email: niessner@tum.de

niessner@tum.de

 

 

 

 

 

Kolloq. Marco Barbina, topic: Implementation scenarios, Technical Challenges and Deep Learning in Autonomous,Avionics Sensors: Research and Future Directions

Thursday 17th of January 2019, 4:15 pm, FMI 03.13.010 (MI-Building, Campus Garching)

 

Abstract:

A suite of Avionic Sensors on board of a modern platform generates an incredible amount of data that needs to be interpreted in a combined way by an expert human operator. In an operative scenario it is useful to adopt a complete suite of sensors like Radar, Visible, IR or Hyperspectral sensors, Transponders, ComInt, etc., on board of a remotely controlled or an autonomous aircraft and in the near future the aggregated bandwidth required to transmit these data towards a ground based operator will soon exceed the capacity of the datalink as well as the dimension and the complexity of the raw data will exceed the time to handle it. This generates the requirement for the elaboration and analysis of the data-set in order to transmit to the ground and present to the operator only the relevant data. Furthermore, data analysis yields poor results unless more sensors are orchestrated and considered as a multi-agent system, each providing a partial view of a complex picture: sensor fusion and distributed awareness are important ingredients for any surveillance systems. The technological challenges of the avionic domain are also dominated by the transition from dedicated and specific HW and CPUs towards processors that have been driving the IoT revolution as well as struggling to find a way to respect the safety requirements of a domain that has only recently accepted software in its single core, single process, single thread. In this talk I will discuss these open research and technical challenges and try to envision a possible scenario for the future.

Bio:

Marco Barbina is Director of Software Engineering in the Airborne and Space Division of Leonardo S.p.A. and manages the teams responsible for the software in several fields ranging from space probes to safety critical and mission critical avionic equipment, from wide band data link and software defined radios to radars, from surveillance to electronic warfare and from flight training simulators to tactical UAVs. He has also been nominated by the Italian Government as member of the group of experts that will define the national strategy for Artificial Intelligence. Marco has received his Master degree in Electronic Engineering at the University of Padua specializing in digital imaging and multilayer video encoding.

contact person:

Dr. Ilias Gerostathopoulos

Phone: +49.89.289.17388

Email: ilias.gerostathopoulos@tum.de

 

 

 

 

 

Kolloq. Nina Schirrmacher, topic: E-Payments in Singaporean Hawker Centers: A Mixed Methods Study

Monday 7th of January 2019, 1:00 pm, FMI 01.13.010 (MI-Building, Campus Garching)

Abstract:

E-Payments in Singaporean Hawker Centers: A Mixed Methods StudySingaporeans are digitally literate, and the country is small and highly connected. The e-payment ecosystem has been flourishing with a multitude of service providers. Yet, the Singaporean e-payment ecosystem is not yet consolidated. While providers face fierce competition in a small market, merchants and customers are hesitant to embrace e-payments. This study therefore investigates the drivers of and barriers to e-payment use in Singaporean hawker centers. Hawker centers portray a microcosm of Singapore’s multicultural society and contribute significantly to the common national identity. 40% of dining occasions take place in 12,000 food stalls across the island, keeping Singaporeans’ living expenses moderate. We use an exploratory sequential mixed methods design based on observations and interviews in the first stage, and structured standardized closed-ended interviews with hawkers in the second stage. Our findings have implications for e-payment providers and government initiatives to improve the e-payment ecosystem.

Bio:

Nina Schirrmacher is a PhD candidate at ESSEC Business School. Her interests cover digital business models, digital platforms and ecosystems strategy, e-payments and Fintech. In her research, Nina investigates how technology enables and impedes the organization of digital ecosystem stakeholders. She holds a double degree M.Sc. Management from the University of Mannheim and ESSEC, as well as an Advanced Master of Research in Business Administration from ESSEC. Nina is currently based at the ESSEC campus in Singapore, where she conducts research on Singaporean food businesses and the Fintech landscape.

 

contact person:

Dr. Manuel Wiesche

Phone: +49.89.289.19539

Email: manuel.wiesche@in.tum.de

 

 

 

 

Kolloq. Prof. Damian Dalton, topic: People don’t buy technology…they are too emotional

Tuesday 18th of December 2018, 11:00 am, FMI 02.09.023 (MI-Building, Campus Garching)

Abstract:

Many startups make the mistake of developing great technology that nobody understands or wants. Conversely, many multinationals have terminated failed projects, the results of which have subsequently be the basis of new highly successful and innovative companies. Regardless of the type of technology or company, the major driving force for any business success is the ability to tune into the emotions and requirements of the market. This lecture reviews the background to successful businesses, particularly technical startups and SMEs, and discusses how companies with limited human and financial resources can use this situation to their advantage, and even compete with, and beat the major behemoths of the technological age. The lecture material will be from my own personal experience in establishing 3 businesses, from my Enterprise and Innovation courses at University College Dublin, and from startup incubator and accelerator sources in Europe and the U.S.

Bio:

Damian Dalton is Assoc Prof. at the School of Computer Science, University College Dublin. He lectures in advanced computer architectures, I.T power modelling, sustainability in data data centres, and enterprise and innovation. He is CEO/CTO of Beeyon, an Irish-based company which has developed and patented, PAPILLON a data centre energy and performance management system. PAPILLON was in the finals of the DataCentreDynamic, Future Technology, international data centre awards in 2015. Beeyon won the 2017 IEEE Irish startup and represented Ireland in Silicon Valley and was the first Irish company to receive funding in Innoenergy’s, the EU largest sustainable energy initiative, Highway programme. He has 5 patents in digital simulation, power modeling and data centre energy management, published over 40 peer-reviewed papers. He is Chairperson and board member of the Irish chapter of the Environmental Association of Universities and Colleges, a board member of the Irish/Swedish Chamber of Commerce and chairperson of the UCD symphony orchestra.

 

contact person:

Thomas Kriechbaumer

Phone: +49.89.289.17678

Email: thomas.kriechbaumer@in.tum.de

 

 

 

 

Kolloq. Prof. Dr. Vitali Gretschko, topic: Sequential Procurement

Thursday, December 6, 2018, at 4:00 pm, FMI 01.10.033 (MI-Building, Campus Garching)

Abstract:

We analyze the problem of a buyer who chooses a supplier for a long-term relationship via auction. The buyer cannot commit that she will not renegotiate the terms of the contract with the chosen supplier in the long run. As the profits in the renegotiation depend on the buyer's information about the suppliers costs, suppliers will be cautious about the information they reveal during the procurement auction. We show that, on one hand, first-price auctions may perform poorly in terms of efficiency and buyer surplus as suppliers may pool on a high bid to conceal information. On the other hand, second-price auctions retain their efficient equilibrium and generate substantial buyer surplus. We demonstrate that, in general, neither first- nor second-price auction are optimal. We derive the optimal mechanism and show that it manages information by concealing the winning bid from the buyer with a positive probability.

Bio:

Vitali Gretschko is head of the ZEW Research Group "Market Design" and Professor of Market Design at the University of Mannheim. He studied mathematics at the University of Münster before working as a business consultant at Accenture. Between 2009 and 2012 Gretschko completed a doctoral degree in economics at the University of Cologne. During this time he spent nine months as a guest researcher at Yale University in Connecticut. Having completed his doctorate, Vitali Gretschko became a postdoctoral researcher at the University of Cologne and worked as a project leader at TWS Partners. Vitali Gretschko's research interests extend across the field of market design, although he is particularly interested in mechanism design, applied auction theory and contract theory.

contact person:

Martin Bichler Phone: +49.89.289.17500

Email: bichler@in.tum.de

 

 

 

 

 

Kolloq. Prof. Blaise Genest, topic: Distribution-based objectives for Markov Decision Processes

Monday, 26th of November 2018, 2:00 pm, FMI 00.12.019 (MI-Building, Campus Garching)

Abstract:

In this talk, we consider a population represented in a fluid/continuous way. A configuration of the population is represented by a distribution over states, and the actions of the controller is represented by a MDP. In this setting, we are interested in considering distribution-based objectives, that is objective talking about the distribution over states (=the configuration of the population). This class of objectives gives rise to an interesting trade-off between full and partial information. As in full observation, the strategy in the MDP can depend on the state of the system, but similar to partial information, the strategy needs to account for all the states at the same time. In this talk, we focus on two safety problems that arise naturally in this context, namely, existential and universal safety. Given an MDP A and a closed and convex polytope H of probability distributions over the states of A, the existential safety problem asks whether there exists some distribution ? i n H and a strategy of A, such that starting from ? and repeatedly applying this strategy keeps the distribution forever in H. The universal safety problem asks whether for all distributions in H, there exists such a strategy of A which keeps the distribution forever in H. Further, we compare these results with existential and universal safety problems for Rabin's probabilistic finite-state automata (PFA), the subclass of Partially Observable MDPs which have zero observation. Compared to MDPs, strategies of PFAs are not state dependent.

Joint work with S. Akshay and Nikhil Vyas.

Bio:

https://perso.crans.org/genest/

 

contact person:

Claudia Link

Phone +49.89.289.17234

Email: link@tum.de

 

 

 

 

14.02.2017

Kolloq. Prof. Kaeli, topic: Expanding the Scope of GPU Computing Through Supporting Diversity

Tuesday, 14th of February 2017, 2:00 pm MI HS 2 (MI-Building, Campus Garching)[mehr]

25.01.2017

Kolloq. Stephan Grell, topic: Bing – Looking at User Data and Processing Tools for Bing Relevance Engineering

Wednesday, 25th of January 2017, 2:15 pm Interims HS 2 (Campus Garching)[mehr]

<< Erste < zurück 1-7 8-14 15-21 22-28 29-30 vor > Letzte >>

Kolloq. Prof. Tobias Höllerer, Ph.D, topic: Simulating the Future of Augmented Reality

Tuesday, 17th of July 2018, 04:00 pm, FMI 00.013.009a (MI-Building, Campus Garching)

Abstract:

AR holds enormous promise as a potentially paradigm-shifting ubiquitous mobile computing technology, enabling the physical world to become the user interface. Clearly, we are still a sizable distance away from this promise, but, just as evidently, many major players in personal computing have already become aware of this potential. In this talk, I will take a look at the features that distinguish AR’s promise from that of other technologies that may also help define the future of personal computing, such as VR, pervasive and ubiquitous computing, agent-based computing, and physical computing. Many of the technologies that are needed for a smooth and seamless AR user experience are still under development. So, how can we help AR along its trajectory and make informed choices about future applications and user interfaces? One track of research in my lab in recent years has been concerned with the simulation of possible future capabilities in AR. With the goal to conduct controlled user studies evaluating technologies that are just not possible yet (such as a truly wide-field-of-view augmented reality display), we turn to high-end VR to simulate, predict, and assess these possible futures. In the very long term, when technological hurdles, such as real-time reconstruction of photorealistic environment models, are removed, VR and AR naturally converge. Until then, we have a very interesting playing field full of technological constraints to have fun with.

Bio:

Tobias Höllerer is Professor of Computer Science at the University of California, Santa Barbara, where he leads the Four Eyes Laboratory, conducting research in the four I's of Imaging, Interaction, and Innovative Interfaces. Dr. Höllerer holds a Diplom in informatics from the Technical University of Berlin as well as an MS and PhD in computer science from Columbia University. He is a recipient of the US National Science Foundation's CAREER award, for his work on "Anywhere Augmentation", enabling mobile computer users to place annotations in 3D space wherever they go. He has been named an ACM Distinguished Scientist in 2013. Dr. Höllerer is author of over 200 peer-reviewed journal, conference, and workshop publications in the areas of augmented and virtual reality, information visualization, 3D displays and interaction, mobile and wearable computing, and social computing. Several of these publications have been selected for Best Paper or Honorable Mention awards at such venues as the IEEE International Symposium on Mixed and Augmented Reality (ISMAR), IEEE Virtual Reality, ACM Virtual Reality Software and Technology, ACM User Interface Software and Technology, ACM MobileHCI, IEEE SocialCom, and IEEE CogSIMA.

 

contact person:

Prof. Gudrun Klinker, Ph.D.

Phone: +49.89.289.18215

Email: klinker@in.tum.de