Max Planck Institute for Informatics

Max Planck Institute for Informatics

Computers, as well as the programs that run on them and the networks they form – with the worldwide Internet leading the way – are the most complex structures ever made by the human beings. This makes computer systems both powerful and mysterious tools. Today's world is a digital world. Ten years ago, data consisted mostly of text; today, however, there is also audio, image and video data. Scientists at the Max Planck Institute for Informatics are concerned with the issue of how we can come to grips with computer systems, and how we can avoid information overload in the modern-day flood of data. The scientists basically want to understand how algorithms and programs work, how complex processes can be simplified, and how we can use the abundance of available data to receive automatic answers from computers to the diverse questions we face.


Campus E1 4
66123 Saarbrücken
Phone: +49 681 9325-0
Fax: +49 681 9325-5719

PhD opportunities

This institute has an International Max Planck Research School (IMPRS):
IMPRS for Computer Science

In addition, there is the possibility of individual doctoral research. Please contact the directors or research group leaders at the Institute.

Department Computational Biology and Applied Algorithmics more
Department Algorithms and Complexity more
Department Computer Vision and Multimodal Computing more
Department Databases and Information Systems more
Max Planck Innovation awards machine text comprehension technology licence more
3D animation with a conventional camera
The 3D movements of a person can be reconstructed based on the recordings of a smartphone or a webcam more
Reality, realistically augmented
A new augmented reality computer programme can edit colours and materials in video streams in realtime more
3D animation – incredibly easy
New software makes it possible to generate animated three-dimensional figures of animals from short videos more
The Max Planck Society met in Saarbrücken
The Max Planck Society's 67th Annual Meeting wrapped up with a panel discussion the "Internet of Things" more
Computer animation: models for facial expression
Computer scientists in Saarbruecken can produce realistic face models, for example for animated films, just from video recordings. more
Onboard computer with a sixth sense
Emergency braking systems already prevent quite a few traffic accidents, but electronic assistants still have no proper overview of what’s happening on the road. Bernt Schiele, Director at the Max Planck Institute for Informatics in Saarbrücken, wants to change this. He teaches computers to anticipate the routes of vehicles and pedestrians. more

In the future, users can take printed multi-touch sensors to create interactive applications  

The piano as a typewriter

The piano as a typewriter

September 18, 2013
Pianists play their instruments as fast as experienced typists on a QWERTY keyboard more
Moveable displays made of paper

Moveable displays made of paper

September 11, 2013
Flexpad turns paper or plastic film into a computer input device and display monitor more
Dolby Laboratories acquires usage rights to innovative imaging patent portfolio from the Max Planck Institute for Informatics more
New keyboard for touchscreens
Computer scientists design new keyboard layout on touch screen devices more
Cebit 2012: 3D animations for everyone
3D movies like "Toy Story" or "Transformers" are based on everyday objects that are able to move like humans. Such 3D characters are created by skilled experts in time-consuming manual work. Computer scientists at the Max Planck Institute for Informatics have now developed two computer programs that can accomplish the same process in mere seconds and can easily be handled even by inexperienced users. more
Bioinformatics and epigenetics – computer-aided cancer diagnosis
Max Planck researchers use ingenious software programs in their quest for biomarkers for clinical cancer diagnosis more
Chasing EHEC with the computer
Scientists in Saarbrücken provide free access to the enteric pathogen’s genetic regulation data more
Researchers normally formulate a hypothesis before beginning an experiment and collecting data. Pauli Miettinen from the Max Planck Institute for Informatics in Saarbrücken is turning this scientific principle on its head with a new procedure for analyzing data – redescription mining. The software can analyze existing datasets and retrospectively extract hypotheses and unexpected correlations. These, in turn, give scientists important clues for asking new questions – for example, when the task is to capture the political mood among the population.
These days, animated figures in films and in computer games are often true to life. After all, they are created with sophisticated three-dimensional models of bodies and faces. Christian Theobalt and his colleagues at the Max Planck Institute for Informatics in Saarbrücken are making it much easier for graphic artists to generate such models – enabling applications that were previously inconceivable.
Movies with audio descriptions help blind people understand the storyline. Could computers take over the task of transforming moving images into natural language? Anna Rohrbach, a scientist at the Max Planck Institute for Informatics in Saarbrücken, and her husband, Marcus Rohrbach, who conducted research at the same Institute until recently, have made it their mission to make that possible. They aim to develop a computer that can automatically generate and read out film descriptions.
His research looks hip and colorful. The prototypes are made from wood, paper and plastic. Cut, printed or pressed. But there’s more to them than meets the eye: Jürgen Steimle and his team at the Max Planck Institute for Informatics and at Saarland University in Saarbrücken are concentrating their efforts on a fully interconnected world in which, for example, computing devices are activated via skin-worn sensors.
Two years ago, countless stories appeared in the media about a computer winning Jeopardy, a tricky quiz show on American television. This raises a number of questions: What can computers know? How do they use this knowledge for language comprehension and for dialog with human beings? And what can be done when machines collect facts about users that aren’t in the users’ best interests?
Emergency braking systems already prevent quite a few traffic accidents, but electronic assistants still have no proper overview of what’s happening on the road. Bernt Schiele, Director at the Max Planck Institute for Informatics in Saarbrücken, wants to change this. He teaches computers to anticipate the routes of vehicles and pedestrians.

Images Take Shape

MPR 4 /2011 Materials & Technology
Humans need only a two-dimensional photo or film to be able to perceive a face or a body in 3-D. Researchers working with Thorsten Thormählen at the Max Planck Institute for Informatics in Saarbrücken are teaching this skill to computers, thus creating new ways of working with images and films.
A computer program helps identify the pathogen responsible for a flu epidemic before it spreads around the globe, giving vaccine developers a valuable head start.
No job offers available

Containing Metastability or: The Art of Choosing between two Equal Bales of Hay

2018 Lenzen, Christoph.; Wiederhake, Ben
Computer Science
Unlike Buridan’s ass, horses do not starve to death when placing them exactly between two identical bales of hay. But what happens, if the horse does not see whether there’s hay to its left and/or right? What if we don’t know either, and if we possibly can’t see where the horse goes? Can we reliably decide whether the horse starves? Interestingly, answers to such questions become highly relevant when considering so-called metastability in circuits. This article explains these terms, their connection, and why it is important to know whether the horse dies or not. more

Reconstruction of cell lineage trees

2017 Andres, Björn1; Schiele, Bernt1; Jug, Florian2; Blasse, Corinna2, Myers, Eugene W.2
Computer Science
Rapid progress in light microscopy allows biologists to image the development of living tissue consisting of several thousand cells. Scientists at the Max Planck Institute for Informatics and the Max Planck Institute for Molecular Cell Biology and Genetics are jointly developing algorithms for reconstructing the lineage of all cells automatically from image data. A central aspect of their research consists in an approach by which two sub-problems are solved jointly, an image segmentation problem and a cell tracking problem. This approach has enabled accurate reconstructions of lineage trees. more
Advances in technologies enable thin and elastic touch displays that can be worn right on human skin. This could help to bring about a new paradigm for interacting with mobile computing devices: human skin turns into an interactive surface, which captures user input and provides visual or haptic output. The long-term goal of this new research direction is to develop more intuitive and more expressive user interfaces for computing devices, which are safe and efficient to use even during challenging mobile activities. more

Solving the Unsolvable

2016 Weidenbach, Christoph
Computer Science
"Learning from Conflicts" is one of the most important strategies for solving computationally hard problems. By guessing a solution that respects local constraints either an overall solution is obtained or a conflict. A conflict can be efficiently turned into further constraints for the problem. "Learning from Conflicts" has pushed the performance of computer programs on hard problems to a new level. For example, automatic verification of computer hardware has turned from an academic discipline into an industry standard. more
Algorithms for knowledge extraction from texts and Web sources have enabled the construction of machine-readable knowledge bases which comprise millions of entities and billions of facts about and relationships between entities. Computers can harness such digital knowledge to semantically interpret natural language, correctly disambiguating even highly ambiguous names and phrases. This deep language understanding opens up better ways of text analytics, question-answer dialogs, and human-computer interaction. more

Distributed Algorithms for Fault-tolerant Hardware

2015 Lenzen, Christoph; Friedrichs, Stephan
Computer Science

The research group "Theory of Distributed Systems" develops algorithms for systems in which many participants collaborate in solving a problem that exceeds the capabilities of each individual participant. Recently, it has been observed that methods and algorithmic ideas from this field facilitate the design of highly robust hardware. Long-term goal of this research avenue is to develop resilient and efficient hardware that is operational even under very harsh conditions such as encountered, e. g., in space.


Scalable learning and perception

2014 Fritz, Mario
Computer Science
While humans handle 10,000 object categories with ease, today's computer vision systems recognize 1,000 categories with moderate accuracy in constraint settings. Our research addresses short comings of today's approaches by focusing on scalable inference that allows for efficient recognition at test time and scalable learning which allows for the acquisition of a large number of concepts. more

Perceptual displays: exceeding physical limitations and improving apparent qualities

2014 Myszkowski, Karol; Ritschel, Tobias
Computer Science

In this report, we focus on the exploitation of perceptual effects to help overcome the physical limitations of display devices in order to enhance apparent image qualities. First, we present apparent display resolution enhancement beyond the physical resolution of display pixels. Then, we discuss various aspects of stereo three-dimensional (S3D) displays that lead to a better control over reproduced depth. Finally, we present an image processing solution, which enables us to see S3D content using any type of stereo glasses and traditional 2D content in glasses-free viewing simultaneously.


YAGO: Knowledge on the Web

2013 Suchanek, Fabian; Hoffart, Johannes; Weikum, Gerhard
Computer Science
YAGO is a large knowledge base, which we built up by information extraction techniques from Internet sources. YAGO contains about 10 million entities and concepts, and knows more than 100 million facts about them. We use techniques from logics to ensure the consistency of our data – YAGO has a false-positive rate of only 5%. YAGO is freely available for download and querying. YAGO is already being used in some semantic projects, and could be the starting point for a new generation of search engines. more

Attacking HIV from new angles

2013 Kalinina, Olga; Pfeifer, Nico
Computer Science

After more than 25 years of research on HIV and AIDS, there is still no approved therapy that would eradicate the virus from an infected patient. At the Max Planck Institute for Informatics, new bioinformatics methods have been developed that, in a patient-specific manner, predict the applicability of a new class of drugs, which block the entry of the virus into human cells. They have investigated the properties of viral and human proteins participating in viral cell entry, and showed how to improve antibody test panels for HIV vaccine development.

Today, there is a vast quantity of knowledge available, in particular on the Internet. In order to precisely answer complex queries in terms of this knowledge, a computer has to be able to perform reasoning on this knowledge. At the Max Planck Institute for Informatics methods are developed that allow a computer to efficiently reason about knowledge bases consisting of several million entries. These methods answer complex queries with respect to such a knowledge base in less than one second. more

Exact Geometric Computing

2012 Sagraloff, Michael
Computer Science
The research group focuses on the development of exact and complete methods to handle complex geometric objects which are fundamental for many geometric algorithms. There exist powerful methods from Algebra to solve the considered problems in theory, however, they turn out to be non-practical because of the high computational cost for the needed symbolic operations. By combining techniques from different mathematical fields, we have been able to develop efficient algorithms which only use a minimum of symbolic operations, whereas most operations are based on fast approximate arithmetic. more

Computer Vision and Multimodal Computing

2011 Schiele, Bernt
Computer Science
The department computer vision and multimodal computing has been founded 2010 with currently 10 scientists. The research areas of the department are on the one hand computer vision with a focus on object recognition and 3D scene understanding and on the other hand multimodal context recognition in the area of ubiquitous and wearable computing. The following summarizes a few representative research themes. more

Optical Performance Capture

2011 Theobalt, Christian
Computer Science
The measurement of human motion (Motion Capture) is an important algorithmic problem in computer vision and computer graphics. At the Max Planck Institute for Informatics we develop new algorithms to reconstruct dynamic models of humans at an unprecedented level of detail from only a handful of multi-view video streams. The reconstructed models consist of detailed 3D geometry, accurate motion data, as well as high resolution surface texture. Our algorithms are not only suitable for dynamic scene reconstruction, but can also be used in other domains, such as advanced video editing. more

Medical Bioinformatics: Molecular Modeling of Diseases and Analysis of Relevant Protein Interactions

2010 Albrecht, Mario; Blankenburg, Hagen; Mayr, Gabriele
Cell Biology Computer Science Medicine
Proteins are involved with many life processes in cells. The knowledge about their complex interplay on the molecular level provides important insights into diseases and cellular mechanisms. Currently, experimentally determined protein interactions are collected worldwide in databases. Therefore, bioinformaticians at the Max Planck Institute for Informatics developed an easy-to-use Internet-based system, which facilitates the global access to the data. Bioinformatics methods are also applied in detailed studies of protein interactions and their 3D structure and disease-relevant function. more

Efficient Querying in Semantic Networks

2010 Neumann, Thomas; Theobald, Martin
Computer Science
Data that exhibits a network or graph structure is ubiquitous, ranging from social networks on the internet to complex networks in life sciences. The combination of network structure within the data and an associated semantic imposes great challenges to searching and automatic reasoning. Our group therefore developed two system for managing large data graphs: RDF-3X, which aims for efficient query processing in classical data graphs, and URDF, which supports searching and reasoning under uncertainty. more

Informatics for meta genome research: insights into the world of non-culturable microorganisms

2009 McHardy, Alice Carolyn; Patil, Kaustubh
Computer Science
Genomic sequencing of microbial communities allows to study the genomes and functional capabilities of uncultivable organisms. At the Max Planck Institute for Informatics methods are developed to assign the obtained genomic sequence fragments to the respective organisms. Application to metagenomes of microbial communities with industrially relevant capabilities, such as the degradation of plant materials to metabolites convertible to biofuels, reveals which organisms have a leading role in these processes. more

Randomized phone chains

2009 Doerr, Benjamin
A phone chain (or phone cascade) is a way of distributing information to all members of a group. Similar problems are subject of research in computer science. It turns out that very good solutions are obtained from phone chains that use randomness. Finding the right dose of randomness is currently a hot topic. more
Motion Capture (MoCap) means to record and analyze human movements. Nearly all commercial products in this field rely on artificial markers, which have to be attached to the person. These markers are usually reconstructed from cameras and used to determine the position and orientation as well as the involved joint angles of the person. In contrast to these reliable marker based tracking systems exist so-called marker-less systems subject to recent research in the fields of Computergraphics, Computer vision and Biomechanics. In this article, we describe current aspects of research which allow for pose tracking of persons without manual intervention, e.g. the use of markers, just by using image data. more

Overcoming Europe's electronic boundaries

2008 Freiheit, Jörn
Computer Science
Europe coalesces. Several challenges in the field of computer science have to be tackled to support the public administrations and their IT systems growing together. The Max-Planck-Institute for Computer Science is involved in one project, funded by the European Commission, which aims at developing and ensuring a secure and efficient collaboration of Europe’s public administrations. Main research topics in this field deal with so-called services that can easily and efficiently be used by citizens and by other administrations with a strong emphasis on protection of data and privacy. more

Computational Epigenetics: Bioinformatics prediction for new approaches to cancer treatment

2007 Bock, Christoph; Lengauer, Thomas
Computer Science Genetics
Damages to the DNA sequence cause only approximately half of all cancers, according to recent research. Equally important – and significantly less understood – are the epigenetic roots of cancer, which involve persistent de-regulation of the human genome. At the Max Planck Institute for Informatics the bioinformatic methods are developed for design and optimization of epigenetic cancer therapies. more

Searching Knowledge instead of Web pages

2007 Suchanek, Fabian; Weikum, Gerhard
Computer Science
How can we make search-engines really "understand" our queries – instead of just finding key-words on Web pages? This article sketches new technologies, which were developed at the Max-Planck-Institute for Computer Science. more

Intelligent Search with Provably Fast Processing Times

2006 Bast, Holger
Computer Science
How can search engines be made more intelligent without sacrificing very fast query processing times? The article sketches a new, interactive search technology that was developed at the Max-Planck-Institute for Informatics, and which addresses these conflicting goals. more

Automatic Theorem Proving in Complex Theories

2006 Sofronie-Stokkermans, Viorica
Computer Science Mathematics
We identify situations in which it is possible to decompose proof tasks in complex theories in a modular way into proof tasks for the simpler components of these theories. As they exploit modularity, such proof procedures are especially flexible and efficient and are therefore widely applicable, for instance in the verification of complex systems, but also in mathematics or knowledge representation. more

Model-Based Animation and Manipulation of Faces in Images

2005 Blanz, Volker; Seidel, Hans-Peter
Based on a vectorspace representation of the shapes and colours (textures) of human faces, which is constructed from a database of three-dimensional (3D) scans, the 3D shape of faces can be reconstructed from single images. Moreover, faces can be animated in 3D or in given images, and facial identity can be exchanged in images, using a general and automated approach. more

Combinatorical optimization

2004 Friedrich Eisenbrand; Markus Behle
Computer Science
Combinatorial Optimization deals with the identification of optimal choices in a large set of alternatives. This topic is of great importance in many fields of applications, such as planning, crew scheduling portfolio management and many more. In this report we give an introduction to combinatorial optimization with the help of a famous example, the traveling salesman problem. Then we describe the integer programming approach to combinatorial optimization and describe two of our results in this area. One concerns the generation of cutting planes and one concerns the complexity of integer programming in fixed dimension. more
Searching and analyzing information in large databases and the World Wide Web has become a key issue for business, society, and the sciences, and it is becoming a bottleneck with the ongoing information explosion. Web search engines often fail to provide good results on particularly challenging expert topics, and the maintenance of thematically dedicated databases can no longer keep up with the rapid data growth. More precise information search requires intelligent methods that incorporate explicit knowledge bases in the form of so-called ontologies and combine them with statistical learning methods. Such approaches have much higher computational costs than the standard techniques used in current search engines. Therefore, fast retrieval of good search results also requires more efficient methods for query processing. This article discusses methods that are being developed at the Max-Planck Institute of Computer Science to improve both the search result quality and the search efficiency. more
Go to Editor View