Association for Computing Machinery
Welcome to the February 1, 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).

HEADLINES AT A GLANCE


Trump Immigration Halt Casts Dark Cloud Over Next IETF Gathering
Network World (01/31/17) Jon Gold

The Internet Engineering Task Force's (IETF) March meeting in Chicago could have a significantly reduced turnout because of the Trump administration's ban on admission into the U.S. of residents of Iran, Iraq, Libya, Somalia, Sudan, Syria, and Yemen. "We now have the situation where people that had planned on attending IETF 98 may not be able to," says IETF member Dan Harkins. Fellow IETF member Joel Snyder notes an inability to attend is a major loss for members, as most Internet standards are characterized by the requests for comment that are the chief business of the task force. "While a great deal of work is done remotely and via email, these face-to-face meetings are jam-packed with technical sessions and discussions," Snyder notes. Iranian-born Norway resident and researcher Naeem Khademi cannot attend IETF 98 because his Iranian passport prevents him from entering the U.S. "I believe as a general principle, IETF should avoid holding meetings in the states that have introduced blanket visa bans on various nationalities (or minorities) indiscriminately, especially if there is an interest within the IETF to have a better participation from such underrepresented countries," Khademi says. The Internet Society's Greg Wood says an IETF subcommittee is working to ensure future meetings are held in as inclusive a setting as possible.


Computer Diagnoses Cataracts as Well as Eye Doctors Can
Live Science (01/31/17) Charles Q. Choi

Scientists at Sun Yat-Sen University in China are using a new artificial neural network called CC-Cruiser to recognize and diagnose congenital cataracts as accurately as eye doctors. CC-Cruiser is a convolutional neural network, which was trained on images from the Childhood Cataract Program of the Chinese Ministry of Health. CC-Cruiser was then given unlabeled data from 57 patients, including 43 with normal eyes and 14 with cataracts. The network identified cases of congenital cataracts with 98.25-percent accuracy, estimated the location of cataracts with 100-percent accuracy, and suggested the proper treatment with 92.86-percent accuracy. In a test using 13 images of normal eyes and 40 pictures of cataracts found on the Internet, CC-Cruiser was slightly less successful. Potential cases of congenital cataracts were identified with 92.45-percent accuracy, location was estimated with 94.87-percent accuracy, and proper treatment was suggested with 89.74-percent accuracy; this decline may have been due to the photographs' variations in lighting, angle, and resolution. In a test comparing CC-Cruiser with ophthalmologists, the network identified all cases with congenital cataracts, while all three eye doctors missed one case. The Sun Yat-Sen researchers say their technology could be adapted to work on other diseases that rely on diagnoses via medical imaging.


Researchers Develop Heat Driven Transistor
R&D Magazine (01/31/17) Kenny Walter

Researchers at Linkoping University in Sweden say they have developed a new transistor that can be used in a wide range of applications, including detecting small temperature differences and using functional medical dressings to monitor the healing process. "We are the first in the world to present a logic circuit, in this case a transistor, that is controlled by a heat signal instead of an electrical signal," says Linkoping professor Xavier Crispin. The researchers say the new transistor is 100 times more sensitive to heat than traditional thermoelectric materials, which means a single connector from the heat-sensitive electrolyte is sufficient. The researchers found one sensor can be combined with one transistor to create a "smart pixel." A matrix of smart pixels then can be used instead of the sensors that are currently used in heat cameras. In addition, the technology could be used to create a new heat camera in a mobile device at a low cost. "When we had shown that the capacitor worked, we started to look for other applications of the new electrolyte," Crispin says.


With or Without a Driver, Vehicles Are Able to Cooperate
Swiss Federal Institute of Technology in Lausanne (01/30/17) Anne-Muriel Brouet

Researchers working on the European AutoNet2030 project are combining driver assistance technologies and inter-vehicle communications to prove convoys of autonomous vehicles will be able to independently operate in high-speed, multi-lane traffic. The proposed method utilizes a distributed communication system and cooperative maneuvering control algorithm developed by researchers at the Swiss Federal Institute of Technology in Lausanne (EPFL). Using the system, connected vehicles in a convoy would be able to communicate directly with other autonomous and manually-driven vehicles in the immediate vicinity. The algorithm uses information collected from the vehicles' sensors to guide the convoy's movements in real time, enabling vehicles to independently adjust their speed and position. Researchers began testing their system by managing robots on simulators before moving to miniature robots and cars on simulators. As part of the AutoNet2030 project, a demonstration was held in October using real vehicles. Researchers equipped a manual car with global-positioning systems, laser sensors, and a human-machine interface to enable the driver to follow instructions from the autonomous vehicles. EPFL's Guillaume Jornod says researchers will be able to improve the multi-lane convoy system once carmakers develop less-expensive solutions for outfitting legacy vehicles with driver assistance technologies.


NCSES Publishes Latest Women, Minorities, and Persons With Disabilities in Science and Engineering Report
National Science Foundation (01/31/17) Robert J. Margetta

The U.S. National Science Foundation's (NSF) National Center for Science and Engineering Statistics (NCSES) on Tuesday announced the publication of its "2017 Women, Minorities, and Persons with Disabilities in Science and Engineering" report, which quantifies the underrepresentation of these populations in science and engineering (S&E). "An important part of fulfilling our mission to further the progress of science is producing current, accurate information about the U.S. STEM (science, technology, engineering, and math) workforce," says NSF director France Cordova. "This report is a valuable resource to the science and engineering policy community." Among the report's key findings is that underrepresented minority women earn a higher proportion of S&E degrees than their male peers. In addition, although women have earned about half of S&E bachelor's degrees since the late 1990s, their representation varies widely by field, ranging from 70 percent in psychology to 18 percent in computer science. Meanwhile, 20 years of progress has not closed the wide gap in educational attainment between underrepresented minorities and whites and Asians. Furthermore, black, Hispanic, and disabled people are underrepresented in the S&E workforce. NCSES also found that two years ago, fewer scientists and engineers were unemployed compared to the general U.S. population.


All In: CMU's Poker Computer Busts Humans Over 20-Day Competition
Pittsburgh Post-Gazette (01/29/17) Sean D. Hamill

A poker-playing algorithm developed at Carnegie Mellon University (CMU) on Monday defeated four world-class human players at the end of a 20-day Heads-Up, No-Limit Texas Hold 'em tournament. Its victory was hailed as "a landmark in [artificial intelligence]" by CMU professor Tuomas Sandholm. He says the win was "awesome" for AI research. The game was the second match between a poker-playing program from CMU and four leading players, who beat the competing algorithm in the first tournament in 2015. AI experts originally predicted it would take up to five years to improve a computer to win a comparable poker match against the best human players. However, the new CMU program, Libratus, achieved victory only 20 months later. Allen Institute for Artificial Intelligence CEO Oren Etizioni says the win "is significant because [poker] is an imperfect information game." Among the enhancements that gave Libratus an edge was a vastly larger number of "core" research hours the CMU team performed on the Bridges computer compared to its predecessor algorithm. "The actual strategy being run on the casino floor [in Libratus] is being run through our compiler," notes Nick Nystrom with the Pittsburgh Supercomputing Center.


A New Approach to 3D Holographic Displays Greatly Improves the Image Quality
KAIST (01/27/17)

Researchers at the Korea Advanced Institute of Science and Technology (KAIST) led by professor YongKeun Park say they have developed a three-dimensional (3D) holographic display whose performance tops that of existing displays by a factor of about 2,600. The researchers say their breakthrough should enhance the limited size and viewing angle of 3D images. The KAIST team used a deformable mirror (DM) in conjunction with two successive holographic diffusers to scatter light in many directions. The results included a wider viewing angle and larger image, along with volume speckle fields caused by the interference of multiple scattered light. The researchers corrected this problem by using a wavefront-shaping method to control the fields. The end-product was an enhanced 3D holographic image with a viewing angle of 35 degrees in a volume of 2 centimeters in length, width, and height. The researchers say the display's performance was more then 2,600 times stronger than the original image definition produced with a DM without a diffuser. "Scattering light has previously been believed to interfere with the recognition of objects, but we have demonstrated that current 3D displays can be improved significantly with an increased viewing angle and image size by properly controlling the scattered light," Park says.


Can an App Help Spies Spot Phony Info? Syracuse Researchers, Others Get $11M to Study
Syracuse.com (01/26/17) Rick Moriarity

Syracuse University researchers are part of a team developing digital tools that will use crowdsourcing and other techniques to help U.S. intelligence analysts determine the credibility of the information they are receiving. The researchers, with colleagues at the University of Arizona, Colorado State University, and not-for-profit SRC Inc., have started work on the project with an $11.5-million contract from the U.S. Intelligence Advanced Research Projects Activity. "Our goal is to create a reasoning and reporting application that is not only effective, but also appealing to users by making the process intriguing and fun while not interfering with their natural reasoning and writing abilities," says Syracuse professor Jennifer Stromer-Galley. The application will guide analysts through a series of steps in which they look at the information they have collected, evaluate the credibility of the information, list the assumptions used in judging the evidence, and identify the information they do not know and determine whether their conclusions might be different if they had the missing information. In addition, the researchers plan to create software that enables intelligence agencies to crowdsource their reports by having groups of people analyze the information they have developed. The researchers also want to use "game-based" principles of human-computer interaction to make the application fun to use.


Intelligence Agency Opens $325,000 Advanced, Automated Fingerprint Gathering Competition
Network World (01/27/17) Michael Cooney

The U.S. Intelligence Advanced Research Projects Activity's (IARPA) newest contest challenges participants to build an automated fingerprint collection system that matches the performance of human operators. The current nail-to-nail (N2N) fingerprint approach requires a trained operator to hold and physically roll the subject's finger over a surface to capture the complete print. The "slap" fingerprint utilizes a single press method that does not require an operator, but the system only captures the parts of the finger touching the sensor. The N2N Fingerprint Challenge offers $325,000 in prizes for devices that can capture the entire fingerprint. Collected data will be compared against N2N and latent data using conventional fingerprint-recognition algorithms, and participants will be judged based on traditional biometric performance measures and the speed of the collection process. The challenge will run in two stages through the fall of this year, ending in a live test in which finalists will demonstrate their devices. The contest is the latest in a series of competitions backed by the U.S. government to spur innovation in the private and philanthropic sectors. Automatic Speech Recognition in Reverberant Environments Challenge, an earlier IARPA contest, asked teams to build speech-recognition systems that can accurately transcribe speech recorded in noisy and reverberant conditions.


Scientists use Microsoft Bands and Bing to Link Lack of Sleep to Cognitive Impairment
GeekWire (01/24/17) Alan Boyle

Microsoft researchers are using wearable fitness trackers and search engine data to study the effect of sleep patterns on cognitive performance. Stanford University Ph.D. student Tim Althoff led the study, which leverages the Microsoft Band, a smartwatch that monitors bodily activity. Data from the Band can be linked to the user's Microsoft account and search history on Bing, and more than 30,000 Band users agreed to have their sleep patterns and their Bing usage data recorded. Microsoft researchers then assessed how quickly users typed in search queries and clicked on links. "Even small differences in the amount of time it would take you to click on the result are indicative of how rapidly you are processing that information," says Stanford professor Jamie Zeitzer. "The idea is people have slower processing speeds as they get more tired." Based on an analysis of 75 million keystrokes and clicks, the researchers found users who slept less than six hours a night for a couple of nights showed signs of cognitive impairment for six days. Microsoft Health's Ryen White says researchers may be able to interpret data about cursor movements and scrolling behavior to spot the early signs of neurodegenerative disorders.


Twitter Data Could Improve Subway Operations During Big Events
University at Buffalo News Center (01/26/17) Cory Nealon

University at Buffalo (UB) researchers have found as subway use grows during events that draw big crowds, so does the number of tweets at those events. The results indicate that data from Twitter, and possible other social media platforms, can be used to improve event planning, route scheduling, crowd regulations, and other subway operations. "Our results show that data from apps like Twitter can help public transportation officials prepare for and react to passenger surges during concerts, baseball games, and other big events," says UB professor Qing He. The researchers conducted the study by gathering subway ridership information from April to October 2014 via turnstiles at Mets-Willets Point station in Queens, NY. The researchers also collected nearly 30 million tweets geotagged to the New York City area during the same time, and then filtered the tweets by the geographic coordinates, the context of the tweet, the time, and other pertinent elements. The researchers used six computer models to analyze the data and found what they call a moderate positive correlation between passenger flow and the rates of tweets during big events. The results indicate increases in social media posts and subway ridership can be linked, and the new method can track this correlation, says UB professor Jing Gao.


Engineering Smart Houses, Driverless Cars
The Cavalier Daily (01/25/17) Katie Lewis

The University of Virginia is building the Link Lab, an interdisciplinary space designed to facilitate studies relating to the implementation of cyber-physical systems (CPS) into daily life. "We are designing a new space, and it will house over 150 researchers in this area," says Link Lab director Kamin Whitehouse. "It's the first space of its kind in the Engineering school, and we think it's the future of engineering." The construction of the lab's physical space should be completed by the end of the calendar year. Whitehouse and his team currently are working on several CPS projects, three of which have won awards from the U.S. National Science Foundation. The Link Lab has three main focus areas of research: wireless health, smart cities, and autonomous robots. Link Lab assistant director Jon Goodall is working on a smart city project involving high-resolution sensing and intelligent algorithms to predict where city flooding will occur and what affect different mitigation efforts might have on the resultant damage. Meanwhile, the lab's autonomous robotics research includes work on intelligent transportation systems. "How do you control, for example, thousands of autonomous vehicles at the same time?" Whitehouse says. "How do you make sure that they won't crash and cause either damage to the equipment or damage to people?"


Is a Master Algorithm the Solution to Our Machine Learning Problems?
TechCrunch (01/30/17) Hassaan Ahmed

Machine-learning programs underpinning search engines, recommender systems, and shopping websites lack a complete "big-picture" understanding of users' habits and preferences, and a "master algorithm" could address this shortcoming, according to Intellisense Solutions co-founder Hassaan Ahmed. He notes many machine-learning experts subscribe to a "connectionist" theory of deducing knowledge via links between neurons, and they see deep learning as the likeliest master algorithm model. Meanwhile, "symbolists" concentrate on philosophy, logic, and psychology, and their machine-learning model emphasizes problem solving using pre-existing knowledge to fill the gaps with an if-then strategy. Other machine-learning schools of thought include "evolutionaries," who base their approach on genetics and evolutionary biology; "Bayesians," who rely on probabilistic inference and Bayes' theorem; and "analogizers," who extrapolate similarity conclusions with a greater focus on psychology and mathematical optimization, using the "Nearest Neighbor" principle. Ahmed says the master algorithm challenge entails designing a single algorithm that can solve all of the problems these schools of thought encompass. Potential pitfalls he envisions in such an achievement include a master algorithm amassing enough knowledge about mankind that it will eventually outsmart humans. Ahmed speculates on scenarios such as super-intelligent machines subjugating or annihilating humanity, or facilitating humans' ultimate evolution.


Abstract News © Copyright 2017 INFORMATION, INC.
Powered by Information, Inc.


To submit feedback about ACM TechNews, contact: [email protected]