Software/Programming

MIT research uses information about how frequently objects are seen together to refine the conclusions of object recognition systems.

by Larry Hardesty

Today, computers can’t reliably identify the objects in digital images. But if they could, they could comb through hours of video for the two or three minutes that a viewer might be interested in, or perform web searches where the search term was an image, not a sequence of words. And of course, object recognition is a prerequisite for the kind of home assistance robot that could execute an order like “Bring me the stapler.” Now, MIT researchers have found a way to improve object recognition systems by using information about context. If the MIT system thinks it’s identified a chair, for instance, it becomes more confident that the rectangular thing nearby is a table.

Read more: Context is ev … well, something, anyway

News

Much scientific research across a range of disciplines tries to find linear approximations of nonlinear behaviors. But what does that mean?

by Larry Hardesty

Spend some time browsing around the web site of MIT’s Computer Science and Artificial Intelligence Laboratory, and you’ll find hundreds if not thousands of documents with titles like “On Modeling Nonlinear Shape-and-Texture Appearance Manifolds” and “Non-linear Drawing systems,” or, on the contrary, titles like “Packrat Parsing: Simple, Powerful, Lazy, Linear Time” and “Linear-Time-Encodable and List-Decodable Codes.”

Read more: Explained : Linear and Nonlinear Systems

Networking

by Larry Hardesty

MIT researchers helped develop a theory that promised much more efficient data networks; then they were the first to put it into practice.

Today, data traveling over the Internet are much like crates of oranges traveling the interstates in the back of a truck. The data are loaded in at one end, unloaded at the other, and nothing much happens to them in between.

Read more: Rethinking Networking

Networking

by Larry Hardesty

A ‘seemingly loopy’ technique that MIT researchers helped develop could dramatically improve the efficiency of communications networks

A radical new approach to the design of communications networks, called “network coding,” promises to make Internet file sharing faster, streaming video more reliable, and cell-phone reception better — among other improvements.

Read more: The Power of 'Random'

Security

A new approach for managing bugs in computer software has been developed by a team led by Prof. George Candea at EPFL. The latest version of Dimmunix, available for free download, enables entire networks of computers to cooperate in order to collectively avoid the manifestations of bugs in software.

A new IT tool, developed by the Dependable Systems Lab at EPFL in Switzerland, called "Dimmunix," enables programs to avoid future recurrences of bugs without any assistance from users or programmers. The approach, termed "failure immunity," starts working the first time a bug occurs -- it saves a signature of the bug, then observes how the computer reacts, and records a trace. When the bug is about to manifest again, Dimmunix uses these traces to rec-ognize the bug and automatically alters the execution so the program continues to run smooth-ly. With Dimmunix, your Web browser learns how to avoid freezing a second time when bugs associated with, for example, plug-ins occur. Going a step further, the latest version uses cloud computing technology to take advantage of networks and thereby inoculating entire communities of computers.

Read more: Safety in Numbers: A Cloud-Based Immune System for Computers

Medical

Scientists have automated the measurement of a vital part of the knee in images with a computer program that performs much faster and just as reliably as humans who interpret the same images.

Having more precise information about wear and tear on this portion of the knee – a blend of fibrous tissue and cartilage called the meniscus – could lead to its use as a biomarker in predicting who is at risk for developing osteoarthritis, researchers say.

Read more: Computers Do Better Than Humans at Measuring Some Radiology Images