Der deutsche Education Blog

Microsoft Research Connections Blog

The Microsoft Research Connections blog shares stories of collaborations with computer scientists at academic and scientific institutions to advance technical innovations in computing, as well as related events, scholarships, and fellowships.

  • Microsoft Research Connections Blog

    New Tools Simplify Data Mining

    • 0 Comments

    In this era of big data, researchers are relying more and more on data mining to help them with their research. Researchers from nearly every field (not to mention businesses from almost every sector) are slicing, dicing, and sifting an exponentially growing mass of data, looking for patterns, trends, and insights. This is powerful stuff, and the essence of the data-intensive “fourth paradigm” of scientific inquiry. 

    Powerful, yes, but also complex. Data mining requires numerous steps: data understanding, data cleaning, model creation, and model comparison. Fortunately, there are new tools for Microsoft Excel that make each step simpler and combine them more seamlessly. 

    New add-ins for Microsoft Excel that simplify data mining are available to download.
    New add-ins for Microsoft Excel that simplify data mining are available to download.

    These tools, collectively known as the Microsoft SQL Server 2012 SP1 Data Mining Add-ins for Office (just rolls off the tongue, yes?) are the product of a joint effort between the Data Mining SQL team and the Microsoft Research Machine Learning and Applied Statistics group.  The tools are available for download.

    Microsoft Data Mining Add-ins help you take advantage of SQL Server predictive analytics in Microsoft Excel and Microsoft Visio. The download includes the following components:

    • Table Analysis Tools for Excel: This add-in provides easy-to-use tasks within Excel that utilize SQL Server 2012 data mining models to deliver quick insights on your spreadsheet data. With just a few clicks, you can analyze key influencers for a given outcome, highlight exceptions, automatically detect groups, fill in missing data, make time series forecasts, or do market basket analysis.
    • Data Mining Client for Excel: By using this add-in, you can create, test, explore, and manage data mining models within Excel, by using either your spreadsheet data or external data that is accessible through your SQL Server 2012 Analysis Services instance.
    • Data Mining Templates for Visio: This add-in enables you to render and share your mining models as annotatable Visio drawings.

    This integrated, comprehensive set of tools should make life simpler for anyone with big data to mine.

    David Heckerman, Distinguished Scientist, Microsoft Research 
    Raman Iyer, Principal Group Manager (Development), SQL Server Business Intelligence, Microsoft Corporation

    Learn More

  • Microsoft Research Connections Blog

    Narwhal Helps Developers Visualize Data in WorldWide Telescope

    • 2 Comments

    First to explain… no, there is no time. Let me sum up: you are a scientist with complex geospatial data visualization challenges. We at Microsoft Research have a solution for you and we’re enhancing this through the release of a software library called Narwhal. (We threw in some example applications as well.) The parent project is Layerscape and the geospatial stories are told by using the WorldWide Telescope visualization engine. The release of Narwhal is in line with our philosophy of “As long as we’re going to build some tools, let’s share them and save others having to re-invent.” Inconceivable! For more: read on!

    Suppose you have some data that you’d like to look at… and it is complicated data. What do I mean by complicated? Perhaps you have a model of an electrical impulse travelling through a maze of 7,000 neurons. Or you have recovered the dive trajectories for the 43 Weddell seals you tagged last summer, or you just derived the magnetic field interactions between Jupiter and Callisto, or the Jaguar supercomputer has finally finished your solution for the thermodynamic structure of the Earth. Let’s run through the two questions that occur to the data visualizer—you—at a time like this: What format should my data be in? And how do I look at it?

    WorldWide Telescope visualization of data on Puget Sound water flow
    WorldWide Telescope visualization of data on Puget Sound water flow

    Unfortunately, there is as yet no single answer to these two questions; and to be fair, you probably already know what format your data is in (be it MATLAB, Comma Separated Value, NetCDF-CF, Microsoft Excel, or whatever). But because your data is complicated, you find it difficult to render and examine on your laptop. Well, we built WorldWide Telescope (WWT) to take advantage of your PC graphics card and now you can look at 500,000 data points as they unfold in time; watch this tour to get the idea. The ability to see the data is just the beginning; we are painfully aware that even though you can see the data, there are lots of other tasks to perform before it is useful, and that is why we built both the Layerscape website (to support content sharing) and the WorldWide Telescope Add-in for Excel (to help you import your data into WWT). All of this you can learn about at Layerscape.

    So far, so good; but if you are really a technical programmer, you will see more potential here—more visualization power—than you can readily access by using Excel. In fact, you may want to be able to connect directly from your software—which helps make sense of your data—to WWT where that data will appear as pixels and lines and circles and polygons and moving sidewalks and drifting balloons and neural impulses and seal-dive trajectories and magnetic fields. Enter Narwhal: software that helps you organize your data and send it to WWT. Narwhal is in its first release, so it is not the ultimate solution, but it does take big jump in that direction. To see what sorts of things Narwhal can help you do, take a look at this video.

    To wrap this up: we are certain that visualization is a key to understanding data, and that humans—and specifically, researchers—are increasingly good at deluging ourselves with massive, complex, hard-to-understand datasets. At Microsoft Research we are both happy and fortunate to get to work on related tools: Layerscape, WorldWide Telescope, and the WWT Add-in for Excel… and now Narwhal. We hope that they find their way to the scientists and educators who need them—and we will continue to refine them, so watch this blog for updates. 

    Rob Fatland, Senior Research Program Manager, Microsoft Research Connections

    Learn More

  • Microsoft Research Connections Blog

    Low-Energy GPS Sensing Looms Large

    • 0 Comments

    Location sensing has become ubiquitous—it’s present every time you turn on your smartphone or engage your car’s navigation system. It’s also become critical to a variety of outdoors and remote research applications, such as wildlife tracking, participatory environmental sensing, and personal health and wellness monitoring.

    The Global Positioning System (GPS) is commonly used for tagging the location of data samples. But traditional GPS location fixing is a power hog; in fact, the typical smartphone battery will drain in about six hours if the phone’s GPS is constantly running, which is particularly problematic in remote locations. Moreover, a smartphone is fairly bulky—not exactly the kind of sensor you can, for example, attach to fruit bats to monitor their nocturnal flights.

    Cloud-offloaded GPS may provide researchers with an energy-efficient solution for location sensing.
    Cloud-offloaded GPS may provide researchers with an energy-efficient solution for
    location sensing.

    In a paper titled, “Energy Efficient GPS Sensing with Cloud Offloading” (PDF file, 6.13 MB), we propose a potential solution to this battery power and size dilemma. This paper describes cloud-offloaded GPS (CO-GPS), an innovative way to perform location sensing by using tiny embedded devices and the cloud to share the work of GPS signal acquisition and processing. By logging only a few milliseconds of raw GPS signals, the device can store enough information for resolving GPS-based location, and it consumes two to three orders of magnitude less energy than stand-alone or mobile phone GPS sensors. The signals are then sent to the cloud with sensor data to reconstruct the location and time that the samples are taken. In delay-tolerant, data acquisition applications—such as animal tracking, float sensor networks, participatory environmental sensing, and long-range time synchronization—CO-GPS is ideal for extending the battery life of mobile devices.

    The paper received the Best Paper Award at ACM SenSys 2012—the premier conference on networked embedded sensing systems and a top forum for the sensor network research community. Many attendees consider the work to be a breakthrough in pushing continuous location sensing to extremely low power devices that can be carried by humans, animals, or recreational equipment.

    We anticipate that CO-GPS will be a boon to citizen-science efforts, particularly those that rely on participatory sensing from embedded devices. For example, the CO-GPS approach is a key enabling technology in Microsoft Research Project CLEO, a participatory environmental sensing system that we are showcasing at the 2012 AGU Fall Meeting this week.

    Jie Liu, Principal Researcher and Research Manager, Microsoft Research, Sensing and Energy Research Group

    Yan Xu, Senior Research Program Manager, Microsoft Research Connections

    Learn More

Page 61 of 128 (384 items) «5960616263»