Welcome to the December 27, 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).

To view "Headlines At A Glance," hit the link labeled "Click here to view this online" found at the top of the page in the html version. The online version now has a button at the top labeled "Show Headlines."

Types of African instruments Computational Study of World Music Outliers Reveals Countries With Distinct Recordings
Queen Mary, University of London
Rupert Marquand
December 21, 2017


Researchers at Queen Mary University of London in the U.K. say they recently conducted the first-ever computational study comparing world music cultures on a large scale. The researchers aimed to identify recordings that have outstanding musical characteristics, called "outliers." The study found that Botswana has the most distinct musical recordings around the world, while China has the most distinct recordings in relation to its neighbors. In addition, the researchers examined geographical patterns of musical outliers for different sets of features and found that Benin has the most outlier recordings with respect to rhythm and harmony, French Guiana has the most outlier recordings with respect to timbre, and Zimbabwe has the most outlier recordings with respect to melody. The researchers used signal-processing tools to extract musical information from audio recordings, data mining to quantify similarity and detect outliers, and spatial statistics to account for geographical correlation.

Full Article
Scientists Develop Method to Track Human Movements More Accurately
EurekAlert
Yan Ou
December 21, 2017


Researchers at the North China Institute of Aerospace Engineering, Hefei University of Technology in China, and the University of North Texas (UNT) have developed a data-driven method to better detect and track human movements for use in a wide range of technologies. The researchers wanted to address the tracking of human subject movement with high accuracy and consistency, going beyond simply tracking a person or a car in a surveillance video or tracking the pose of a person to estimate their actions. They used a time-of-flight camera and a three-dimensional (3D) point cloud to identify five extreme points on the human body; after those points are marked, then the joints can be identified. "Given that our depth-imaging device can acquire only surface data of a 3D volume, a detected extreme point could become invisible after an action, [such as] rotation, which makes the consistency of detection critical," says UNT professor Xiaohui Yuan.

Full Article

People exchanging money behind curtain Scientists Learned to Predict Public Corruption With Neural Networks
Higher School of Economics (Russia)
December 21, 2017


Researchers at the Higher School of Economics (HSE) in Russia and the University of Valladolid in Spain have developed a neural network prediction model of corruption based on economic and political factors. The researchers used a database combining the main cases of political corruption in Spain, proposing an early warning corruption model to predict whether corruption cases are likely to arise in Spanish regions given specific macroeconomic and political factors. The researchers used self-organizing maps, a neural network method, to predict corruption cases over different timeframes. The results show economic factors are relevant predictors of corruption, including the taxation of real estate, economic growth, increased house prices, and the growing number of deposit institutions and non-financial firms. The researchers also found the same ruling party remaining in power too long is positively related to public corruption.

Full Article
SUTD Researchers Discover a Valleytronics Route Towards Reversible Computer
Singapore University of Technology and Design
December 21, 2017


Researchers at the Singapore University of Technology and Design (SUTD) have developed a versatile all-electric-controlled valley filter and detailed for the first time a solid working design of a valleytronic logic gate that executes the full set of two-input Boolean logics. The traditional model for generating a logically reversible computer depends on complex circuitries that unavoidably produce large quantities of wasteful bits. The key distinction of the SUTD team's proposed valleytronic-based reversible logic gate is its retention of additional bits of input data in the valley state of the computational output to support logical-reversibility, circumventing complex circuitries and reducing the generation of wasteful bits. The researchers say this architecture also is more compatible with the expanding industrial and commercial demands for compact smart devices with contracting form factors. "The union of valleytronics, digital information processing, and reversible computing may provide a new paradigm towards the future of ultimately energy-efficient computers with novel functionalities," says SUTD professor Ricky Ang.

Full Article

Brown dog lying down Hacked Dog Pics Can Play Tricks on Computer Vision AI
IEEE Spectrum
Jeremy Hsu
December 22, 2017


Researchers at the Massachusetts Institute of Technology (MIT) have demonstrated a new way to fool computer vision algorithms that enable artificial intelligence systems to see. The researchers exploited the Google Cloud Vision API that enables anyone to perform image labeling, face and landmark detection, optical character recognition, and tagging of explicit content. Traditional hacking approaches are inefficient and impractical when targeting large images with tens of thousands of pixels. To overcome this problem, the MIT team adapted a "natural evolution strategies" method that generates smaller populations of images around the larger image, with large random groups of pixels being perturbed instead of single pixels. Then, given the classifier's output on these randomly perturbed images, the system recovers what the contribution of each individual pixel is to the classification output, according to MIT researcher Andrew Ilyas. The researchers used this method to create "adversarial images" that would trick a computer vision program into seeing an object that was not really there.

Full Article
The AI That Can Choose the Perfect Picture
Daily Mail (United Kingdom)
December 22, 2017


Google researchers have developed a neural image assessment system (NIMA) that uses artificial intelligence (AI) to scan a group of photos and then help the user pick the most attractive ones. The system uses deep convolutional neural networks to scan photos for both technical and aesthetic elements. The researchers want to develop the new system as an app to suggest improvements such as changes to brightness and contrast in real time. The app also could offer tips to improve framing and the "aesthetic beauty and emotional appeal" of images, according to the researchers. They developed the system using data based on what human judges generally select as good images in photo contests. An algorithm then rates photos on technical elements such as blurriness, highlights, and the use of shadows. The algorithm gives all photos a composite score from one to 10, and suggests edits to improve each photo's score.

Full Article

Autonomous cars driving through intersection with illustrated sensors around them New Depth Sensors Could Be Sensitive Enough for Self-Driving Cars
MIT News
Larry Hardesty
December 21, 2017


Researchers at the Massachusetts Institute of Technology (MIT) Media Lab are developing imaging systems that utilize "time of flight," an approach that measures distance by gauging the time it takes light projected into a scene to bounce back to a sensor. The researchers say the new approach to time of flight imaging increases the technology's depth resolution 1,000-fold, a breakthrough that could make self-driving cars practical. In addition, they say the new approach could enable accurate distance measurements through fog, which has proven to be a major obstacle to the development of self-driving cars. At a range of two meters, existing time of flight systems have a depth resolution of about a centimeter. However, the new MIT system has a depth resolution of three micrometers, suggesting that at a range of 500 meters, the system could achieve a depth resolution of only one centimeter.

Full Article
Watch Robots Do Chin-Ups, Push-Ups, and Sit-Ups for the Sake of Science
The Los Angeles Times
Amina Khan
December 20, 2017


A team of Japanese researchers has developed Kengoro and Kenshiro, two humanoid robots that can perform push-ups, do crunches, stretch, and even sweat while exercising. Because of the robots' ability to perform human-like movements, they could serve as models to help scientists better understand how the human body moves. "Our intent is to design a humanoid based on human systems--including the musculoskeletal structure, sensory nervous system, and methods of information processing in the brain," the researchers say. The team designed Kengoro and Kenshiro using human statistical data to give the robots more human-like proportions in terms of their mass distribution and the size of each body part. Kenshiro and Kengoro are tendon-driven robots and have 114 joint degrees of freedom, compared to 548 joint degrees of freedom for humans, and just 55 joint degrees of freedom for standard axial-driven humanoid robots.

Full Article

Marilyne Andersen and Siobhan Rockcastle, the developers of OCUVIS New Software Can Model Natural Light From the Occupants' Perspective
Swiss Federal Institute of Technology in Lausanne
Cécilia Carron
December 20, 2017


Researchers at the Swiss Federal Institute of Technology in Lausanne (EPFL) in Switzerland say they have developed OCUVIS, visualization software that enables architects to simulate three-dimensional (3D) building models to assess the performance of natural light indoors. After specifying the ambient conditions, architects can use OCUVIS to view the visual and non-visual characteristics of the resulting natural light in their designs. A 3D computer model of a planned building and its geographic position initially is used to help specify the ambient weather conditions and see how the resulting daylight would be perceived by building occupants. An architect can use OCUVIS to compute and visualize the results, and change settings related to the time of day or year, sky condition, and location in the building to see how the dynamics of sunlight, weather, and space impact building occupants. In addition, the EPFL researchers have developed a visualization platform and computational models describing the underlying performance factors.

Full Article
Metal Printing Offers Low-Cost Way to Make Flexible, Stretchable Electronics
NC State News
Matt Shipman
December 20, 2017


Researchers at North Carolina State University (NCSU) have developed a method for directly printing metal circuits, creating flexible, stretchable electronics. The method uses multiple metals and substrates, and is compatible with existing manufacturing systems that utilize direct printing technologies. The technique is based on existing electrohydrodynamic printing technology, which already is employed in many manufacturing processes that use functional inks. However, instead of ink, the NCSU researchers use molten metal alloys with melting points as low as 60 degrees Celsius. The researchers demonstrated the new method using three different alloys, printing on four different substrates--one glass, one paper, and two stretchable polymers. During testing, the researchers examined the resilience of the circuits on a polymer substrate and found the circuit's conductivity was unaffected even after being bent 1,000 times. In addition, the team found the circuits were still electrically stable even when stretched to 70 percent of tensile strain.

Full Article
Novel Computational Approach Launches New Paradigm in Electronic Structure Theory
Michigan State University
December 20, 2017


Researchers at Michigan State University (MSU) focusing on quantum calculations have proposed a new computational model for solving the Schrodinger equation that holds the key to defining the molecular and atomic motion of electrons. The researchers say their method offers a new paradigm for obtaining highly accurate electronic energies by integrating deterministic coupled-cluster and stochastic Quantum Monte Carlo strategies. The technique exhibits rapid convergence toward target molecular electronic energetics based on the data derived from the early stages of the Monte Carlo wave function propagations, cutting computational costs by orders of magnitude. The stochastic methods are employed to identify the leading wave function components while the deterministic coupled-cluster computations, combined with suitable energy corrections, provide the missing information.

Full Article
New AI Method Keeps Data Private
University of Helsinki
Minna Merilainen-Tenhu
December 20, 2017


Researchers at the University of Helsinki and Aalto University in Finland, along with colleagues at Waseda University of Tokyo in Japan, say they have developed a new machine-learning method that is based on the concept of differential privacy, which guarantees the published model or result can reveal only limited information on each data subject. "Previously, you needed one party with unrestricted access to all the data," says University of Helsinki professor Antti Honkela. "Our new method enables learning accurate models, for example, using data on user devices without the need to reveal private information to any outsider." As an example, Honkela notes the new method is capable of analyzing cell phone data or predicting cancer drug efficacy using gene expression while also ensuring the privacy of the data subjects. "Learning from big data is easier, but now we can also get results from smaller data," says Aalto University professor Samuel Kaski.

Full Article
New Quick-Learning Neural Network Powered by Memristors
The Michigan Engineer News Center
Dan Newman
December 19, 2017


Researchers at the University of Michigan (U-M) say they have invented a memristor-assembled reservoir computing system that reduces training time and enhances the capacity of similar neural networks. They note the use of memristors lowers space requirements and enables easier integration within existing silicon-based electronics. The team employed a special memristor that retains events only in the near history, enabling them to bypass an expensive training process while still providing the network with recall capability. When the reservoir is fed a dataset, it identifies important time-related features of the data, and passes it to a second network in a simple format, so the second network only needs training like simpler neural networks, changing weights of the features and outputs that the first network handed it until it reaches an acceptable level of error. "The beauty of reservoir computing is that while we design it, we don't have to train it," says U-M professor Wei Lu.

Full Article
Discover Lifelong Learning Resources
 
ACM Publications
 

Association for Computing Machinery

2 Penn Plaza, Suite 701
New York, NY 10121-0701
1-800-342-6626
(U.S./Canada)



ACM Media Sales

If you are interested in advertising in ACM TechNews or other ACM publications, please contact ACM Media Sales or (212) 626-0686, or visit ACM Media for more information.

To submit feedback about ACM TechNews, contact: technews@hq.acm.org