July 12, 2010 July 13, 2010

Faculty Summit 2010

Location: Redmond, WA, USA

  • SONGO improves the mobile-search user experience by caching common search results and web pages on mobile devices. First, a community-based cache is created by mining the most popular queries in mobile-search logs. This cache is updated daily, making sure that the latest popular information is always available locally. Over time, the community-based cache is personalized by adding all user search queries that do not exist in the cache. An analysis of four months of mobile-search logs shows that, on average, approximately 70 percent of search queries submitted by a user can be answered by caching 2,500 links on a 2 MB cache. A prototype implementation in Windows Mobile demonstrates responses 18 times faster and offering 30 times more energy savings comparing with querying through a 3G link.

  • LittleRock is a new mobile-phone-sensing architecture that enables continuous sensing by offloading low-level sensing tasks to a low-power, embedded processor. Mobile phones are ideal platforms for developing human-centric applications that adapt to the user, because the sensors on phones can track user context. Although there are already many applications that use such sensors, state-of-the-art phone architectures do not support continuous sensing to determine full user context. With LittleRock, energy-efficient continuous sensing on mobile phones will enable true user-centric applications, resulting in a much richer user experience. We demoed continuous audio monitoring using LittleRock, which enables scenarios such as conversation post tagging and audience-aware reminders.

  • Hawaii is a project to explore the combination of mobile devices with cloud services. We are assembling a set of cloud services that might be useful for creating mobile applications. These services include computation, storage, location, notification, and identification. We also plan to build an application exchange using these services. We are providing access to the above set of services, also known as the Hawaii platform, to several computer-science classes at top universities. The pilot offering of these classes occurred in spring 2010.

  • Microsoft has innovated continually in developing novel interaction modalities, or natural user interfaces. Surface and Project Natal are two examples. While these modalities rely on sensors and devices situated in the environment, we believe there is a need for new modalities that enhance the mobile experience. We take advantage of sensing technologies that enable us to decode the signals generated by the body. We demoed muscle-computer interfaces, electromyography-based armbands that sense muscular activation directly to infer finger gestures on surfaces and in free space, and bio-acoustic interfaces—mechanical sensors on the body that enable us to turn the entire body into a tap-based input device.

  • Gustav is a realistic painting-system prototype that enables artists to become immersed in the digital painting experience. A natural interface makes Gustav ideal for hobbyists and professional artists alike. Gustav achieves a high level of interactivity and realism by leveraging the computing power of modern GPUs, taking full advantage of multitouch and tablet input technology and our novel, natural media-modeling and brush-simulation algorithms. Our prototype provides convincingly realistic models for pastel and oil media, with more to come.

  • Microsoft Research’s Project Salzburg can generate a procedural model from a single audio clip and apply perturbations at variation from a single model. This means it can change sound files according to what happens in a game world at a certain moment, delivering an amazing experience.

  • Microsoft Research Software Radio (Sora) is a novel software-radio platform with full programmability on commodity-PC architectures. Sora combines the performance and fidelity of hardware-based software-defined-radio (SDR) platforms with the programmability and flexibility of general-purpose processor SDR platforms. Sora uses both hardware and software techniques to address the challenges of using PC architectures for high-speed SDRs. The Microsoft Research Software Radio academic kit is available for research purposes.

  • WorldWide Telescope now features the largest digital-image mosaic of Mars ever created: 13,000 HiRISE images, with guided tours of Mars narrated by NASA scientists Carol Stoker and Jim Garvin.

  • Terapixel provides the largest, clearest image of the sky. We showcased a wide range of Microsoft technologies: .NET parallel extensions, Dryad, Trident, and high-performance computing.

  • See the best student design solutions related to this year’s theme of “Service meets Social” from the Art Center College of Design, Carnegie Mellon University, Central Saint Martins College of Art and Design, New York University’s Interactive Telecommunications Program, Universidad Iberoamericana, and the University of Washington.

  • Discover ways to connect students with internships and jobs through the Imagine Cup, a student technology competition challenging students to tackle current real-world problems while strengthening technical, problem-solving, and communication skills that can translate into a future career. Learn about being a mentor to an Imagine Cup team. Also, discover Students to Business (S2B), a program that connects students with Microsoft partners and customers to gain practical experience for entry-level positions and internships. S2B students benefit from unique training and certification opportunities to build the skills they need for the job market.

  • We presented a pair of demos from world finalists for the Touch and Tablet Accessibility Track for Microsoft’s 2010 Imagine Cup.

    • OneView is a Tablet PC-based application that enables students, both blind and sighted, to create, read, and edit diagrams collaboratively. OneView provides a synchronized multimodal interface—featuring visuals, audio, and text—that enables students to use their preferred interface mode while collaborating with other students.
    • Note-Taker is a portable, custom-designed hardware/software assistive device that improves the accessibility of higher education for students who are legally blind or have reduced vision. The technology enables such users to view streaming video of a classroom presentation while simultaneously taking notes in a split-screen interface with Microsoft OneNote.
  • As video and still cameras have become almost ubiquitous, people are taking increasingly more photographs and videos. Often, the photographer’s intent is to capture more than what can be seen in a single photograph, and he or she instead takes a large set of images or a video clip to capture a large scene or a moment that extends over time. One can combine these images to produce an output that improves the input images, such as creating an image with a large field of view, a panorama, or a composite image that takes the best parts of the image: a photo montage. But creating these results is still non-trivial for many users. One challenge is in creating large-scale panoramas, for which the capture and stitching times can be long. In addition, when using consumer-level point-and-shoot cameras and camera phones, artifacts such as motion blur appear. Another challenge is combining large image sets from photos or videos to produce results that use the best parts of the images to create an enhanced photograph. We presented several new technologies that advance the state of the art in these areas and create improved user experiences. For panorama generation, we demonstrated ICE 2.0, stitching of panoramas from video, and generating sharp panoramas from blurry videos. For generating composites, we demonstrated video to snapshots and de-noising and sharpening using lucky imaging.

  • The Boogie intermediate verification language is a simple, imperative language with two important attributes. First, its expressive syntax and precise semantics enable natural and unambiguous encoding of the operational semantics of many real-world imperative programming languages, such as x86, C, C#, and Java. Second, it is equipped with a verifier that compactly encodes the operational semantics of a Boogie program into a logical formula whose satisfiability can be checked using powerful solvers, such as Z3. These attributes make the Boogie system useful for building a variety of symbolic program-analysis tools. We showcased four static-analysis tools with different goals, all built on top of Boogie:

    • Dafny, for proving full functional correctness of C#-like object-oriented programs
    • VCC, for proving full functional correctness of C concurrent programs
    • STORM, for finding violations of code-level protocols in concurrent programs
    • SymbolicDiff, for static differencing of program behaviors of evolving C/C# programs.
  • Tracing just-in-time compilers (TJITs) determine frequently executed traces—hot paths and loops—in running programs and focus their optimization by emitting optimized machine code specialized to these traces. Prior work has established this strategy to be especially beneficial for dynamic languages such as JavaScript, in which the TJIT interfaces with the interpreter and produces machine code from the JavaScript trace. We try to address these issues in the SPUR research project, in which we are developing a prototype TJIT for Microsoft Common Intermediate Language (CIL, the target language of C#, VisualBasic, F#, and many other languages). Working on CIL enables TJIT optimizations for any program compiled to this platform.

  • The Revisions project introduces a novel programming model for concurrent applications. Our prototype provides C# programmers with a simple yet powerful and efficient way to share data safely among concurrent tasks, using concurrent revisions and isolation types. The Task Parallel Library (TPL) is a domain-specific embedded language for expressing concurrency through the use of first-class anonymous functions (delegates) and parametric polymorphism (generics). It has been integrated into Microsoft .NET 4.0 and ships with Microsoft Visual Studio 2010.

  • FORMULA (Formal Modeling Using Logic Analysis) analyzes domain-specific modeling languages (DSMLs) and model transformations that are expressed in a logico-algebraic framework. For example, the “domain” of a DSML is characterized by a term algebra and a set of invariants over sets of terms. Similarly, model transformations can be expressed as axioms that map sets of terms from one domain to sets of terms of another. FORMULA uses these descriptions to calculate models that satisfy—or do not satisfy—these axioms. This basic capability can be used to determine properties of DSMLs/transformations such as domain equivalence, domain emptiness, and structure-preserving maps.

  • CHESS is a tool for finding and reproducing heisenbugs in concurrent programs. It repeatedly runs a concurrent test, ensuring that every run takes a different interleaving. If an interleaving results in an error, CHESS can reproduce the interleaving for improved debugging. CHESS is available for both managed and native programs. Cuzz is an effective tool for finding concurrency bugs. It works on unmodified executables and is designed for maximizing concurrency coverage for your existing, unmodified tests. It randomizes the thread schedules in a systematic, disciplined way, using an algorithm that provides probabilistic coverage guarantees. Cuzz is scalable and can run on large programs that create lots of threads.

  • Performance measurement of software is a critical component of software development. Performance traditionally has been measured using profiling, which is often too little (only certain inputs are profiled) or too late (to make requisite changes to address the root cause before shipping). The SPEED project attempts to address these limitations by static estimation of the symbolic computational complexity of programs. It builds over recent advances in static program analysis, which traditionally has been used for checking correctness, as opposed to measuring performance.

  • Code Contracts provide a language-agnostic way to express coding assumptions in Microsoft .NET programs. The contracts take the form of preconditions, postconditions, and object invariants. Contracts improve program robustness and reliability. We showed how to use the contracts for runtime checking, static contract verification, and automatic documentation generation. Code Contracts can be used with all .NET programming languages for teaching and as a platform for research.

  • Pex and Moles are Microsoft Visual Studio 2010 Power Tools that assist in unit testing of Microsoft .NET Framework applications.

    • Pex automatically generates test suites with high code coverage. Right from the Visual Studio code editor, Pex finds interesting input-output values of your methods, which you can save as a small test suite with high code coverage.
    • Moles enables you to replace any .NET method with a delegate. Moles supports unit testing by providing isolation via detours and stubs. The Moles framework is provided with Pex and can be installed by itself as a Visual Studio add-in.
  • Microsoft Translator is the web-scale machine-translation service run by Microsoft Research, featuring a rich API and collaborative features that enable developers to combine human edits with machine translation intelligently. The API extends the capability to integrate automatic translation capabilities deeply into an application to any developer, for any research project, similar to what Microsoft has shown in Internet Explorer, Office, Bing, and Instant Messenger.

  • The Chemistry Add-in for Word makes it easier for students, chemists, and researchers to insert and modify chemical information, such as labels, formulas, and 2-D depictions, from within Microsoft Office Word. In addition to authoring functionality, Chemistry Add-in for Word enables user denotation of inline “chemical zones,” the rendering of high-quality and print-ready visual depictions of chemical structures, and the ability to store and expose semantic-rich chemical information in a semantically rich manner.

  • Zentity, Microsoft Research’s research-output-repository platform, aims to provide the necessary building blocks, tools, and services for developers creating and maintaining an organization’s repository ecosystem.

  • Microsoft Academic Search is a free academic search engine developed by Microsoft Research. It provides many innovative ways to explore scientific papers, conferences, journals, and authors, connecting millions of scholars, students, librarians, and other users.

  • We showed how Bing product search, returning object-level information for the matching product only, can be enriched by enabling navigation to related products. For each related product, our demo visualizes the type and strength of its relationship to a matching product, which complements existing “black-box” recommender systems showing recommended items with no proper explanation. Technical challenges include disambiguating different types of relationships and comparing the strength of relationships across types, which normally would require domain expertise.

  • We presented an interactive multimedia-exhibition system for showing ancient Chinese paintings of moving focus. We use one of the most famous Chinese paintings, Along the River During the Ching-ming Festival, as the exhibit. The master piece is scanned into a high-resolution, gigapixel image. The system not only presents the details of the painting, but also uses hundreds of voice dubbings and environmental sounds, organizing them into short stories to reveal the prosperous daily life of people in the Song Dynasty.

  • The Microsoft Academic Search Kiosk is built on the Microsoft Academic Search platform. It enables users to see basic information about scholars attending the Microsoft Research Faculty Summit, such as number of papers and citations, and the network of authors.

  • The Microsoft Biology Foundation (MBF) is a language-neutral bioinformatics tool kit built as an extension to the Microsoft .NET Framework. It implements a range of parsers for common bioinformatics file formats; a range of algorithms for manipulating DNA, RNA, and protein sequences; and a set of connectors to biological web services such as the National Center for Biotechnology Information’s BLAST. MBF is available under an open-source license, and executables, source code, demo applications, and documentation are freely downloadable.

  • Scientific applications have diverse data and computation needs that scale from desktop to supercomputers. Besides the nature of the application and the domain, the resource needs for the applications also vary over time—as the collaboration and the data collections expand, or when seasonal campaigns are undertaken. Cloud computing offers a scalable, economic, on-demand model well-matched to evolving eScience needs. We presented a suite of science applications that leverage the capabilities of the Windows Azure cloud-computing platform. We showed tools and patterns we have developed to use the cloud effectively for solving problems in genomics, environmental science, and oceanography, covering both data- and compute-intensive applications.

  • Researchers need access to large-scale, real-world data when performing data-driven research at web scale. N-gram models, a type of probabilistic model for predicting the next item (word or token) in a sequence, can be used to further research in the areas of search (query suggestion and clustering), information retrieval and extraction, machine translation, speech, spelling, learning, and others. Microsoft Research, in partnership with Bing in the Online Services Division and in collaboration with professors worldwide, is providing researchers with access to petabytes of data in the form of n-grams. This cloud-based platform delivers as many as 5 grams on new content types such as title and anchor texts, in addition to document body and new model types such as smoothed models via on-demand services and regular updates. Web N-gram services are enabling new research directions and repeatability of experimentation via data benchmarks.

  • Space debris has become increasingly problematic, as was made clear by the collision between the satellites Cosmos 2251 and Iridium 33 in February 2009. To track the remnants of such events, a combination of observation, data management, modeling, and operational tools requires burst-compute capability. As a global problem, this requires a global IT infrastructure. The Southampton Space Situational Awareness Azure System provides processing and visualization of tracked space objects and conjunction/impact analysis. It also supplies a decision-support system for space debris and near-Earth-object impacts.

  • Introducing Kinect for Xbox 360, a revolutionary new way to play: no controller required. See a ball? Kick it, hit it, trap it, or catch it. If you know how to move your hands, shake your hips, or speak, you and your friends can jump into the fun—the only experience needed is life experience.

DemoFest Posters

  • Client + Services + Cloud for Cancer Bioinformatics
  • Computational Biology Application Suite for High Performance Computing
  • Development and Application of Wireless Network of Geosensors for Environmental Monitoring in Tropical Forests
  • Digital Urban Informatics: Computational Innovations for Sustainable Urban Environmental and Water Resources Management
  • e-Farms: a 2-way Road from Small Farms into the Networked World
  • LEAD II: Hybrid Workflows in Atmospheric Science
  • MapReduce and Clouds for Science
  • Multi-Touch Human Robot Interaction for Disaster Response
  • Multi-Touch Interfaces for Content-Based Video Searches
  • Putting the Surface in Context
  • Statistical Methods for Large Data Sets in Search and Advancement