In this list:
Computer Science projects
Earth, Energy, and Environment projects
Education and Scholarly Communication projects
Health and Well-being projects
Computer Science
See also:
Microsoft Research INRIA Joint Centre
BSC – Microsoft Research Centre at Barcelona Supercomputing Centre
Reliable and Efficient Concurrent Object-Oriented Programs (RECOOP)
Bertrand Meyer, ETH Zurich, Switzerland
Socially Structured User Behavior and Externalities in Sponsored Search Auctions
Sander Bohte, Nicole Immorlica, Vangelis Markakis, Han La Poutre
Center for Math and Computer Science, Netherlands
AdRules: Improving Quality of Ads
Krzysztof Dembczynski, Wojciech Kotlowski, Dawid Weiss
Poznan University of Technology, Poland
Web-scale Semantic Social Mash-Ups with Provenance
Harry Halpin, Henry Thompson, University of Edinburgh, United Kingdom
Goal-Driven Information Retrieval
Jun Wang,
University College London, United Kingdom
Parallel MatLab and Financial Algorithms
Anne Trefethen, Mike Giles
Oxford e-Research Centre, Oxford, UK
The aim of this project has been to create science services for the water research community to archive and perform simple analyses of data. Oxford installed, run and tuned MatLab software on the Microsoft clusters.
HPC Tools for the Automotive Industry
Michael Resch
HLRS, University of Stuttgart, Stuttgart, Germany
Stuttgart University High Performance Computing (IHR) group is leading in research in engineering together with industry and at the same time is an outstanding center for high performance computing. This is a unique combination worldwide to demonstrate the value of Microsoft solutions in the HPC space and for mission critical applications in world leading car manufactures (Mercedes and Porsche).
Stuttgart has a unique setting in the field of engineering simulation. The institute of High Performance Computing (IHR) is closely attached to the High Performance Computing Center Stuttgart (HLRS) and is itself part of the school of engineering. This supports a very interdisciplinary approach. Stuttgart itself is a centre for automotive industry with DaimlerChrysler and Porsche as well as suppliers like Bosch. IHR has established close collaborations with these companies and is hence in a unique position to undertake the suggested project work. As the lead researcher is at the same time the director of HLRS he has well established links to the operational side of simulation at the companies mentioned. This puts the researchers as close as possible to industrial usage of simulation as soon as developed and made available.
Earth, Energy, and Environment
See also: Microsoft Research - University of Trento Centre for Computational and Systems Biology
Climate Induced Vegetation Change Analysis Tool (CLIVT)
Evgeny Loupian, Mikhail Zhizhin
Space Research Institute (IKI) and Geophysical Center (GC), Moscow, Russia
Space Research Institute (IKI) and Geophysical Center (GC) are the key players in the collection of remote sensing, environmental and geophysical data in the Russian Academy of Sciences. IKI has a strong research group working on applications of remote sensing to study boreal ecosistems with a unique collection of wide-area vegetation indices and other biophyscial parameters derived from the sattelite observations. GC has a unique technology for parallel data mining and change detection in the very large climate data archives. This is an opportunity to apply high-performance computing and Microsoft technology to study the regional climate change impact on various ecosystems on the vast territory of Norhtern Eurasia.
FENS (Fighting Emissions with Nuclear Magnetic Resonance)
Lynn Gladden
University of Cambridge, Magnetic Resonance Research Centre (MRRC), UK
As identified by the forward-look document “Towards 2020 Science”, the need for new approaches to energy research is of increasing importance for economic and social development. With this project, the Magnetic Resonance Research Centre (MRRC) seeks to establish a research collaboration with Microsoft to develop the next generation of magnetic resonance (MR) techniques to be applied in the field of chemical engineering research, with a particular focus on developing new measurement capabilities which will provide insights for the design of processes which reduce energy demand – identified as the most effective means of reducing carbon dioxide (CO2) emissions, since the energy ‘not used’ reduces carbon flows all the way down the energy supply chain and conserves the hydrocarbon resource. Researchers at MRRC currently lead the world in their ability to apply ultra-fast magnetic resonance imaging (MRI) techniques to engineering (i.e. non-medical) systems; this ability to ‘look inside’ chemical processing operations can transform our understanding of how chemical processes operate. In many cases MR is the only technique available which can image within a 3-D optically opaque system, with chemical specificity. This project aims to develop new data acquisition strategies which, we anticipate, will reduce data acquisition times by at least an order of magnitude (down to 1 ms) and may allow to distinguish structural features that are inaccessible to measurement using current MR methods. In particular, the aim is to characterise the dimensions of microstructures that exist within rapidly time-varying environments, a typical example might be measuring the instantaneous size of a bubble in a turbulent two-phase flow. This will be achieved by developing new data sampling strategies and image reconstruction algorithms, characterised by a move away from traditional signal processing to Bayesian signal processing methodologies. This will provide new information upon which to base theoretical and numerical models of the complex processes occurring within multi-component, multi-phase, reactive systems which characterise the chemical process industry. It is another step towards achieving chemical process design based on robust, scientific understanding which is essential for safe, sustainable process development and operation. The work proposed in this project addresses applications in the chemical processes industries in which an increase in MRRC’s ability to re-design these processes and/or improve their operation will lead to a decrease in CO2 emissions of 6000 million tonnes per annum. This is approximately equal to the annual CO2 production of the United States.
Climateprediction.net
Myles Allen
Oxford University, Oxford, UK
The climateprediction.net project, through our ongoing collaboration with the BBC, is currently running a large ensemble simulation of climate change from 1920-2080 using coupled atmosphere-ocean general circulation models running on a volunteer network of personal computers.
In collaboration with Microsoft the aim is to complete the coupled experiment, demonstrate the value of its outputs, archive the data and make it available to the climate research community, with particular emphasis on making probabilistic climate forecast products accessible to developing countries
Swiss Experiment for Environmental Studies
Karl Aberer
EPFL, Lausanne, Switzerland
The natural environment is undergoing dramatic changes, yet all too often we cannot provide satisfactory answers to open questions, such as: "how much change is anticipated?" and "what are the main causes and consequences of such change?" A prominent example of such change is global warming, which strongly influences alpine ecosystem and hydrologic function as well as the formation of hazards from alpine peaks to valley bottoms. The primary limitation to address these socially relevant questions has been the essential lack of appropriate spatial and temporal environmental observations across the landscape in which environmental engineers and scientists can test and validate models which simulate future scenarios and make real time predictions (e.g. flooding, debris flows, droughts, rock falls/avalanches).
The Swiss Experiment (“SE”) project addresses this issue by developing a large-scale environmental sensing approach deploying 'classical' and new generation instruments from different disciplines in an interdisciplinary collaborative effort.The SE is a large scale research effort funded by the Swiss National Science foundation through the NCCR-MICS research centre and by the ETH council through the CCES research centre.
This collaboration will be focused on supporting and enhancing cutting-edge environmental science research through the application of emerging sensor network technologies. In particular, it will leverage the MSR SensorMap platform and other appropriate Microsoft technology, Institute technology and tools, and third party technologies, to enable the publishing, archiving, visualization, and analysis of environmental data from: (1) a pilot sensor network deployment on the EPFL campus, "Sensorscope" (initially); (2) a large scale deployment in the Swiss Alps, "Swiss Experiment" (later). SensorMap will provide support for archiving, publishing, querying and analyzing real-time as well as archived sensor data over a geo-centric web interface.
Microsoft High Performance Computing Technologies for Safety and Environment
Yuriy Boldyrev, Alexander Snegirev, Sergey Lupuleac
Saint Petersburg University, St. Petersburg, Russia
This two-core project aims to implement HPC technologies by Microsoft into the real life engineering problems of fire safety and environmental protection at the prestigious Polytechnic University of Saint Petersburg.
The first sub-project (FireEx) will develop a new efficient computational methodology capable of predicting the interaction of fine water spray with buoyant turbulent diffusion flame occurring in fires. Windows-based parallelized software tool will be developed that will allow the regimes of flame extinguishment to be quantitatively investigated and the optimum droplet size to be determined for a given fire scenario. Experience in the use of Microsoft platform for massive parallel computations in solving the resource-consuming engineering problem will be gained for further migration to Windows-based parallel computer technologies. Due to its multi-disciplinary nature, this sub-project will include both theoretical (spray and flame model developments) and heavily loaded computational (massive parallel simulations of turbulent reacting drop-laden flows using Large Eddy Simulation technique) components. As a result of this sub-project, a challenging and long-standing engineering problem will be addressed: industrial designers of water mist fire extinguishing systems and fire engineers will be able to apply the knowledge obtained in this project to create more efficient equipment for fire suppression, yet keeping the advantages of acceptable cost and low (or no) damage to material and the environment. The results of this sub-project will be disseminated throughout the domain community and mass media.
The second sub-project (Dam) is focused on detailed mathematical simulation of the dynamic behavior of the gates of Saint Petersburg flood defense system under the loads from flowing water. Accurate time dependent CFD (computational fluid dynamics) simulations are needed to determine the structure of the flow around the gate section. The dynamic behavior of the gate section within submerging process is to be modeled by the methods of fluid structure interaction on the basis of obtained CFD results. It is planned to perform serial simulations to consider different regimes of operation (different depths, flow conditions etc.). Such simulations need the massive parallel computations on the multiprocessor computers. It is planned to use the MS CCS cluster of the Laboratory of Applied Mathematics and Mechanics as well as world leading commercial CFD codes Ansys Fluent and Ansys CFX for these simulations. On the base of obtained numerical results it is planned to produce the data base. This data base will be implemented further for simulation of the gate submerging process.
Regional Climate Change Projections for Southern Africa
Bruce Hewitson
University of Cape Town, South Africa
The project will deliver a framework for exploring and an initial assessment of the envelope of projected regional climate change for southern Africa. The tool and methods developed will then be made available to the broader community for expanded experiments using distributed computing to complement the continued in-house activities. This project addresses the two key limitations on climate change science in southern Africa; a constrained capacity to undertake robust exploration of the range of possible future climate at a regional scale, and the limited understanding of the complex interaction of key physical processes governing the regional climate response to anthropogenic forcing.
A high resolution physically based modeling system will be used to allow for a regional assessment of the envelope of future climate change at high spatial and temporal resolution. This will be achieved through coupling a coarse resolution global and a high resolution regional climate model and through new approaches to analyzing the complex signal to establish the robust regional change signals. The models involved are the UK Met Office Hadley Centre’s global coupled model HadCM3 and regional model HadRM3. Both are leading models in their fields and have respectively been made available to run on PCs through the ClimatePrediction.net (CPDN) and PRECIS (Providing Regional Climates for Impacts Studies) projects. The project extends the work of the CPDN project and focus on the critically vulnerable southern Africa region where current capacity does not yet allow for an analysis of this project’s scope.
In parallel to the core modeling activity the project will build on preliminary work at the University of Cape Town which looks at maximizing the information content of the ensemble output within the complex mix of deterministic, stochastic and non-linear responses characteristic of a quasi-chaotic system. This component stands to significantly enhance the value of the project, as well as bringing greater value to the broader communities’ activities on model-based climate simulation.
Climate Modeling
Richard Jones
The Met Office Hadley Centre, Oxford, UK
The project will deliver a physically based modeling system allowing detailed assessments of future climate change for any region by continuously coupling coarse resolution global and a high resolution regional climate models. The models involved are the Met Office Hadley Centre’s global coupled model HadCM3 and regional model HadRM3. Both are leading models in their fields and have respectively been made available to run on PCs through the ClimatePrediction.net (CPDN) and PRECIS (Providing Regional Climates for Impacts Studies) projects. UK NERC has funded prototype Windows and Linux systems which allow standard PCs to run and couple the basic models at the heart of CPDN and PRECIS. The system that this project will deliver extends the CPDN project, which enables the running of large ensembles of models to explore the range of likely global climate change, by adding the high resolution model HadRM3. The large-ensemble high resolution regional climate modeling capability that it will enable is the key to providing an assessment of likely ranges of detailed future climate changes over any region of the globe. The resulting system will be ideally suited for deployment on multi-processor or multi-core PC systems or via public resource distributed computing.
For the different experiments that will be designed by the collaborating regional projects public volunteer computing will be a key resource with the planned use of multi-processor and multi-core computers including dedicated powerful compute servers a powerful addition to this. For the system to deliver the range of information required using public computers it is essential that it can be run under Microsoft operating systems. Using this full range of resources to run state of the art global and regional climate models will demonstrate the capabilities of Microsoft technology. It will also provide valuable experience in their capabilities to deliver a range of outcomes in a key field of scientific computation.
Virtual Fire
Kostas Kalabokidis
The University of the Aegean, Mytilene, Greece
This project, a Web GIS platform for forest fire management based on Microsoft Virtual Earth TM will be developed in order to share the data and information produced by the GoND Lab easily, validly, promptly and to multiple end-users. The use of a Web-based platform will eliminate the need to install special desktop software. This cuts down on client deployment time and costs to zero, and enables any authorized user to immediately access the platform from anywhere in the world. Users will have the ability, without the requirement of knowing the handling of commercial and complicated GIS applications, to utilize the capabilities of GIS; to query on the databases of GoND Lab and to immediately receive answers; to locate points of interest; to connect their Palmtop or their GPS with the platform; and to download information provided by the administrator of the system.
Four Remote Automatic Weather Stations and a Weather Forecasting System based on SKIRON and RAMS limited area weather forecasting systems will provide all the necessary data for fire prevention and early warning, channeled through the platform to the end-users. Additionally, a geographical representation of the fire risk potential and identification of high-risk areas at different local regions will be provided on-demand, based on meteorological (wind speed, fuel moisture, precipitation, etc.), socio-economic (distances from urban areas, roads, power lines, waste disposal areas, etc.) and biophysical (cover types, fuel models, terrain, etc.) parameters. By using the FARSITE and BehavePlus software, maps will be produced that will graphically represent the spread and intensity of a forest fire at different times and spaces. By using these models, the ability to users will be granted (forest fire fighting personnel, emergency crews, authorities, etc.) in order to design an operation plan to encompass the forest fire, choosing the best way to put out the fire with the proper means at the proper time, etc.
Furthermore, the system may optionally provide in the future the ability to its endusers to locate on-line and in real-time vehicles of the Fire Service and other resources by using GPS systems and communications that will transmit the coordinates of each item to the system, portraying them on an electronic map. With this future extension the dissuasion of traffic jams and mismanagement will be achieved, effacing their negative consequences to the extinguishments procedures and achieving better coordination in resource dispatching. Thus, the whole scheme would be functioning as an integrated early warning and decision support system.
Education and Scholarly Communication
FaceBots: Robots Utilizing and Publishing Social Information in FaceBook
Nikolaos Mavridis, Tamer Rabie
United Arab Emirates University, UAE
Health and Well-being
See also: Microsoft Research - University of Trento Centre for Computational and Systems Biology
Machine Learning Tools for Genomics & Genetic Research
Richard Durbin
Wellcome Trust Sanger Institute, Cambridge, UK
This collaboration will bring together a world-leading centre for genomics and bioinformatics with one of the world’s strongest industry groups in statistical machine learning.
Common diseases such as cardiovascular disease, cancer, obesity, diabetes and psychiatric illnesses are caused by a combination of multiple genetic and environmental factors. Understanding how the genetic factors interact with each other and with the environment would allow better prevention, diagnosis and treatment of these diseases, and thus allow individualized treatment of these diseases based on the genetic make-up of the patients.
Whilst previous studies have focused on single diseases, the Wellcome Trust Sanger Institute is tackling the much more ambitious goal of identifying all common diseases that can be reliably predicted from markers across the entire genome. This would allow diagnosis of a large range of diseases at an early stage, with potentially huge impact on management and treatment of these diseases.
Almost all existing approaches for studying the genetic causes of disease are localized to studying effects of a very small set of genes, and thus are not capable of capturing subtle effects of many genes.
The Wellcome Trust Sanger Institute proposes a genome-level computational analysis that integrates genetic and functional genomic data to study effects of multiple genes jointly. This would be a novel approach and have substantial scientific impact.
An interesting addition to this research is a parallel probabilistic inference system tied into the Infer.NET framework which will be developed by Thore Graepel and Ralf Herbrich (both at MSRC). This will allow biological models written for Infer.NET to be parallelised automatically and tackle more complex models without the (large) investment of development time in parallelising each and every algorithm used.
Being able to analyse more data using more complex models could allow us to discover more subtle relationships between genome and disease. Hence, the project could potentially have significant positive effect on the outcome of the Sanger project. In any case, it would allow greater throughput of experiments and hence more effective use of the resources allocated to the project.
Cancer and Cardiac Image Analysis
Mike Brady
Oxford University, Oxford, UK
More people in the Western World, both women and men, are killed by cardiovascular disease than any other disease; according to recent figures, it annually accounts for almost 1 million deaths in the USA and over 4.3 million deaths in Europe (AHA 2005,BHF 2005).
The recent advent of “real-time” 3D ultrasound in the commercial market which uses a matrix-array transducer to acquire volume time-series of ultrasound data ie 4D (3D+time) data, is attracting significant interest in clinical cardiology. It can be used both for diagnosing the health of the heart at rest and when it is under stress (either by exercise or pharmacologically) - so-called stress echocardiography. The “holy grail” in stress echocardiography is to be able to automatically segment such sequences to extract a patient-specific model of the health of the heart. This project aims to look at this problem.
We propose to investigate graph-cut based methods as the basis for spatio-temporal segmentation of 3D+time ultrasound sequences of the heart. This research will further the understanding of how graph-cut methods can be applied to solve spatio-temporal problems as well as potentially offer a new way to automatically derive quantitative 3D information about the health of the heart in a robust and efficient way. The latter would pave the way for cardiologists to make more informed decisions about disease assessment and treatment monitoring on the millions of people world-wide who have cardiac conditions.
SPUMS (Saving Patients Using Microsoft)
Frederique Lisacek, Ron Appel
Swiss Institute of Bioinformatics, Geneva, Switzerland
A recent study identified hospital care as the third leading cause of deaths in the US. One reason for these deaths is interactions between recently prescribed drugs and the drugs currently in the patients system. The total aim of the project is using Mass Spectrometry to systematically screen the patients for drugs, present in the patients system, and to use a reference database to warn of possible conflicts with other prescribed drugs.
Microsoft’s involvement in the project is to assist in the development of a generic data management framework. This framework will consist of a central database of detectable chemical fragments with their associate mass spectra. The development of a process to capture, translate and compare mass spectra extracted from patients with the reference database. To enable equivalent World Wide projects to utilize this data management framework and to allow the synchronization of their library of Mass Spectra knowledge with the ones created by SIB.
HPC Tools for Genetic Studies in Neurological Disorders at the Western Galilee Hospital
Assaf Schuster
Technion, Haifa, Israel
The specific goal of this research is to develop new methods and advanced software to analyze SNP data on larger families than currently possible, and to apply the new tools on some neurological disorders at the Western Galilee Hospital. Understanding the genetic makeup of this disease, for which several genes are already known, will strengthen the analysis of common neurological diseases such as Parkinson. The project will develop several methods, including improved HMM algorithms, variational methods with bounds, variational sampling and a distributed environment for genetics and made them available to practicing geneticists around the world. The graphical models and techniques so developed will fit well the current trends of systems biology in which diverse set of data types are integrated to study a wide set of biological phenomena. The use of graphical models will allow incorporating expression data as a specification of phenotypes, as well as knowledge about gene ontology and protein structure and protein function databases.
The software will be deployed in a production system to support genetic analysis. The system consists of a hierarchy of resource pools, managed by status of the art grid software. The largest pools are expected to contain tens of thousands of compute resources; mainly consisting of MS Windows based PCs and MS-CCS PC clusters.
Applications of Survey Propagation Algorithms to Computational Biology for the Cure of Multiple Diseases
Riccardo Zecchina
Politecnico di Torino, Turin, Italy
Main scope of the project is to attack one of the most challenging reverse engineering problem of computational biology, namely the use of the DNA chip data to reconstruct the gene-to-gene interaction network. The MSR research technology on which this application will be built is a new family of combinatorial optimization algorithms which have been developed recently in the context of the study of critically constrained hard optimization problems. These algorithms have a probabilistic nature and are extremely efficient, two facts which make them ideal candidates for dealing with noisy reverse engineering problems. It is a well-established fact that the technology to be used outperforms all known algorithms on benchmarks problems.
Computational biology is a very important and active field of modern science in which we expect our technology to give important contributions: The potential payoff from reconstructing gene-regulatory networks from DNA micro-array data is enormous since these networks are expected to provide the key information for understanding, controlling and finally curing many human diseases caused by genetic mis-regulation, and to suggest possible new drug targets. The biggest problem is, however, the inversion of DNA micro-array data to infer genetic interactions, and our techniques are arguably among most promising candidates in the world for achieving this data inversion.
Moreover, the reconstruction methods could be extended to networks that are active in a coupled way, such as the host and parasite networks during the infection cycle. This would open a Pandora’s box of possibilities in the cure of parasitic infectious diseases such as malaria.
There is also a second application in the project which will funded by a partner institution. It deals with applying the above-mentioned MSR technology to another important problem in computational biology, more precisely in computational neuroscience, namely to reconstruct synaptic learning mechanisms in neural networks which are both experimentally verifiable and computationally efficient.
Any progress in the neuroscience part of the project would have crucial consequences, e.g. it could lead to potential understanding and control of Alzheimer’s disease, or to mechanisms for re-establishing muscular control of people with spinal injuries.
HPC Tools for High Throughput Analysis of Complex Genetic Data
Iain Buchan
University of Manchester (NIBHI), Manchester, UK
In October 2006 Microsoft, NIBHI in Manchester and partners ran a successful proof of concept showing that Microsoft technologies and development tools can be used to enable the parallelisation of statistical analyses by research teams without specialist HPC knowledge, thereby bringing the computational development closer to the statistical and genomic thinking. Microsoft portal and workflow technologies, linked to compute clusters, also facilitated the sharing of evolving analyses and workflows within multi-disciplinary teams.
This project proposes to develop the proof of concept into a stable platform with sufficient functionality for effective application by a range of research groups. We propose developing the platform around the needs of researchers looking at HIV treatment, anticoagulation, rheumatoid arthritis (RA), asthma, type 2 diabetes and obesity. The aim is to build a computing platform, accessible to all relevant disciplines, for exploring the complex genetic determinants of disease. The approach is based on the application of ‘slices’ through the Microsoft technology stack, and relevant e-science solutions to deliver a component of the ‘Social BioHealth e-Laboratory’ specifically to support medical genetics high-throughput analyses.
