Welcome to the March 22, 2017 edition of ACM TechNews, providing timely information for IT professionals three times a week.
ACM TechNews mobile apps are available for Android phones and tablets (click here) and for iPhones (click here) and iPads (click here).
|
A Robotic Revolution in Healthcare
BBC News
Kenneth Macdonald
March 20, 2017
Researchers at the Edinburgh Center for Robotics, a joint initiative between Heriot-Watt University and the University of Edinburgh in the U.K., are using artificial intelligence (AI) to create robots that will learn from their environment, each other, and humans. In addition, the U.K.'s Engineering and Physical Sciences Research Council is funding the development of four new robots that will use AI to transform healthcare and emergency room response. Heriot-Watt professor Katrin Lohan says a new era of robotics will require contributions from computing, engineering, and the social sciences. "We are interested in how to develop robots that are programmable by everyone, so that they can learn from us in natural interaction rather than us just sitting there and typing in code," Lohan says. Edinburgh University professor Sethu Vijayakumar says robots will transform how people live and work over the next two decades.
|
DARPA Wants to Cultivate the Ultimate Transistor of the Future
Network World
Michael Cooney
March 17, 2017
The U.S. Defense Advanced Research Projects Agency (DARPA) this month will present the Dynamic Range-enhanced Electronics and Materials (DREaM) program, which aims to develop a new generation of radiofrequency (RF) and millimeter-wave transistors to address the power and range requirements for wireless devices in a range of applications. DARPA program manager Dan Green says DREaM transistors will transmit and receive the large and complex RF signals of the future, and they will do so in smaller packages while consuming minimal power. DREaM researchers will try to solve the performance tradeoffs between four key characteristics of RF transistors, including signal power, power efficiency, the range of frequencies in which the transistors work, and the measure of the fidelity at which a receiver can amplify signals. In addition, DREaM will focus on developing new materials that can handle more electrical charge and voltage without degrading.
|
PRACE Proceeds Into Second Phase of Partnership
HPC Wire
March 20, 2017
Members of the Partnership for Advanced Computing in Europe (PRACE) this month ratified a resolution to proceed with PRACE 2, the second phase of their partnership, which is designed to strengthen Europe's position as a leading scientific supercomputing provider. The goal of PRACE is to provide a federated European supercomputing infrastructure that is science-driven and globally competitive. The new PRACE 2 program will help to create a basis for the sustainability of the infrastructure, in order to continue fostering world-class science as well as enabling technology development and industrial competitiveness in Europe through supercomputing. The project will involve the provisioning of a federated world-class Tier-0 supercomputing infrastructure, which is architecturally diverse and allows for capability allocations that are competitive with other programs in the U.S. and Asia. In addition, PRACE 2 will institute a thorough peer-review process, coordinated high-level support teams, and implementation actions in dissemination, industry collaboration, and training.
|
Using Big Data to Analyze Images and Video Better Than the Human Brain
Uni Research
March 20, 2017
Researchers at Uni Research's Center for Big Data Analysis in Norway are developing advanced methods for using big data in image analysis and recognition. "Our researchers have developed specialized knowledge about handling huge amounts of data, and thus how essential knowledge can be identified," says Uni's Eirik Thorsnes. He thinks traffic safety improvement is the area in which image analysis currently has the most potential. Uni researcher Alla Sapronova trains computers in a child-like way, showing them patterns of input signals and telling them what to expect the output signal to be, repeating the process until pattern recognition is achieved. "Then I show the computer an input signal, such as an image, that it has not seen before and test whether the system understands what it is," Sapronova says.
|
MIT-Stanford Project Uses LLVM to Break Big Data Bottlenecks
InfoWorld
Serdar Yegulalp
March 20, 2017
Researchers at the Massachusetts Institute of Technology's Computer Science and Artificial Intelligence Laboratory and Stanford University's InfoLab have developed Weld, a common runtime for data analytics that takes the disjointed pieces of a modern data processing stack and optimizes them in concert. Weld, which is written in the Rust programming language, generates code for an entire data analysis workflow that runs efficiently in parallel using the LLVM compiler framework. Each piece of the data stack runs fast, but data movement across the different functions can dominate execution time, according to the researchers. They say Weld addresses this problem by creating a runtime that each library can plug into, providing a common method to run key data cross the pipeline that needs parallelization and optimization. Efforts such as Weld, with no existing dependencies, are likely to become the standard-bearers for the Rust language as it matures.
|
Research Teaches Machines to Decipher the Dawn Chorus
Engineering and Physical Sciences Research Council
March 20, 2017
The U.K. Engineering and Physical Sciences Research Council is supporting research on deciphering the timing and sequences of bird calls, led by scientists at Queen Mary University of London (QMUL). The work builds on the Warblr smartphone app, which can analyze recorded bird sounds and match them with patterns of bird calls in its dataset and list possible corresponding species. The latest research will enhance the analysis to learn what birds are communicating and which birds dominate the conversations. "We use modern machine learning methods where you don't necessarily know how a computer has made a decision about a particular sound, but by training it we can encourage a computer algorithm to generalize from those," says QMUL's Dan Stowell. The researchers' longer-term goals include achieving better understanding of other animal groups' social structure, and how human language evolved. Stowell also envisions applications involving audio as well as visual analysis.
|
Molecular Motor-Powered Biocomputers
Dresden University of Technology (Germany)
Stefan Diez
March 20, 2017
European researchers recently launched a research project that aims to develop a biocomputer based on highly efficient molecular motors that will use a fraction of the energy of existing computers, and that can address problems where many solutions need to be explored simultaneously. The project will use biomolecular motors as computing units, in which each nano-sized machine can solve problems by moving through a nanofabricated network of channels designed to represent a mathematical algorithm, an approach known as "network-based biocomputation." When the biomolecules reach a junction in the network, they either add a number to the sum they are calculating or leave it out, which makes each biomolecule act as a tiny computer with a processor and memory. The researchers will focus on developing the technology required to scale up network-based biocomputers to a point at which they can compete with other alternative computing approaches.
|
Coders and Librarians Team Up to Save Scientific Data
Computerworld
Sharon Gaudin
March 20, 2017
Software programmers, librarians, and others are archiving scientific data from government websites, especially with the Trump administration proposing deep budget cuts to agencies that support significant research, as well as government-sanctioned deletion of important data. "We're most concerned that data might be taken offline and public accessibility will be gone and it'll only be available as [Freedom of Information Act] requests," says University of Pennsylvania librarian Margaret Janz. "Our goal is to make trustworthy copies of data so it will be available to the public and suitable for research." Janz helps organize archiving events via the DataRefuge program, whose volunteers only copy data in the public domain. The process starts with the nomination of URLs for storage in the nonprofit public Internet Archive, with more complex data "harvested" by participants using scripts and tools developed with either the R or Python coding languages.
|
Making Better Decisions When Outcomes Are Uncertain
MIT News
Larry Hardesty
March 21, 2017
Researchers at the Massachusetts Institute of Technology and Duke University have demonstrated the possibility of building accurate Markov decision processes (MDPs) while gathering less empirical data than had previously been necessary. Their algorithm accomplished this by applying the median of means, a statistical technique that divides a sample of random values into subgroups, extracts the average or mean of each of those, and then yields the median of the outcomes. The team showed that, in the context of straight averaging, the number of samples needed to estimate the mean value of a decision is proportional to the square of the range of values that the value function can assume. The median of means makes the number of samples proportional to the much narrower range of the Bellman operator. MDP analysis is designed to ascertain a set of policies, or actions under specific conditions, that maximize the value of a reward function.
|
It Begins: Bots Are Learning to Chat in Their Own Language
Wired
Cade Metz
March 16, 2017
OpenAI's Igor Mordatch is investigating new ways for machines to converse with each other by building virtual environments in which software bots learn to generate their own language out of necessity. The two-dimensional environments enable the bots to create a language to collaborate and complete assigned tasks via reinforcement learning. The bots construct language by assigning random abstract characters to simple concepts they learn through navigation, including each other. Mordatch says these techniques can help machines better understand language and enable a true conversational interface. Mordatch and colleagues note their approach diverges from most artificial intelligence research, which attempts to mimic human language using deep neural networks. "For agents to intelligently interact with humans, simply capturing the statistical patterns is insufficient," the researchers say. "An agent possesses an understanding of language when it can use language (along with other tools such as non-verbal communication or physical acts) to accomplish goals in its environment."
|
Hand Movement, No Fraud
National Research Nuclear University
March 14, 2017
Researchers at the National Research Nuclear University's Institute of Cyber Intelligence Systems in Russia have developed InCallAuth, a mobile application that enables a smartphone to recognize its owner by a characteristic movement of the hand when answering a call. The application uses data from the phone's accelerometer, gyroscope, and light sensor, while the initial device position, the speed of the movement of the hand holding the smartphone, and the change of the phone's position in space are used as parameters. If the application does not recognize the owner, it will ask for a password, without which it would be impossible to answer an incoming call. The researchers say authentication via hand movements and gait is the future of mobile devices because it is very difficult to counterfeit these parameters.
|
Waiting to Be Sold: IUPUI Researchers Develop Model to Predict Probability of Home Sales
IUPUI Newsroom
Richard Schneider
March 14, 2017
Researchers at Indiana University-Purdue University Indianapolis (IUPUI) have developed machine learning algorithms that can predict how long it will take for a house to sell. The algorithms are based on methodology used to predict the length of disease survival in patients with life-threatening medical conditions. The algorithms account for how changing features, such as lowering the price of the home or adding a bathroom, influences the length of time the house is for sale. The researchers trained the algorithms with three months' worth of data on 7,216 houses in five Indiana cities. The data included details typically found in real estate listings, as well as the dates of the initial listing and the sale, enabling the system to study features and patterns. IUPUI professor Mohammad Al Hasan notes the methodology can be used to examine other regions with different real estate factors and predict when a home will sell.
|
How Aristotle Created the Computer
The Atlantic
Chris Dixon
March 20, 2017
Modern computers would not exist without the influence of Aristotelean principles. For example, philosopher-mathematician George Boole credited Aristotle's Organon, which set forth logical structure, schema, and laws, for inspiring his work, "Laws of Thought," to give Aristotelean logic a precise algebraic notation. Claude Shannon built on Boole's work to propose a governing systematic theory for the design of electrical circuits, and a framework for building arithmetical logic units. Alan Turing later developed a template for computer design based on mathematical logic in an attempt to find an algorithm that solved the "decision problem" for determining the truth or falsehood of an arbitrary mathematical statement. The revelation that such an algorithm could not exist led to a mathematical model of an all-purpose computing machine, and later evidence of how a program could be stored within a computer alongside the data upon which it operates.
|
|
