Welcome to the March 20, 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).
|
Researchers Are Using Darwin's Theories to Evolve AI, So Only the Strongest Algorithms Survive
Quartz
Dave Gershgorn
March 20, 2017
Researchers at Google Brain and OpenAI are applying Darwinian principles of evolution to advance artificial intelligence (AI). Google's neuroevolution project trained 1,000 image-recognition algorithms on deep-neural networks to recognize specific images. The more accurate algorithms were then copied and "mutated" to see if their clones' accuracy would improve, with such mutations allowed to survive and eventually achieve 94.6-percent recognition accuracy. Meanwhile, OpenAI's research focused on using "worker" algorithms to train a master AI to perform an unknown task. The evolutionary AI tracks how workers learn, thus learning how to extract more insight from the same amount of data. The workers played Atari and reported their scores to the master. The highest-scoring algorithms were copied and randomly mutated, then put back into rotation so subsequent mutations could be copied or deleted, depending on their scoring prowess. OpenAI's approach is considered to be closer to evolution's true biological function.
|
NTU Singapore Collaborates With Chinese & Japanese Institutions
Asian Scientist
March 20, 2017
Researchers at Nanyang Technological University, Singapore (NTU Singapore) have entered into a research agreement with colleagues from the University of Chinese Academy of Sciences (UCAS) and the RIKEN Institute in Japan to enhance joint research collaborations, as well as to increase visits and exchanges for students and academic and research staff. The researchers will focus on translating existing and future research advances in artificial intelligence, the Internet of Things, big data analytics, and virtual reality into useful commercial products and services. For example, the NTU-University of British Columbia Research Center of Excellence in Active Living for the Elderly will work with UCAS to develop potential applications in healthy aging and education. Meanwhile, the partnership with RIKEN aims to leverage both institutions' scientific expertise and research facilities to establish joint research centers in Singapore and Japan. Initial projects will focus on mind/body interaction and understanding various diseases, particularly during the pre-clinical stage.
|
X-rays Map the 3D Interior of Integrated Circuits
IEEE Spectrum
Rachel Courtland
March 17, 2017
Researchers at the Paul Scherrer Institute in Switzerland have used x-rays to peer within commercial integrated circuits and reconstruct them in three dimensions. The team used x-ray ptychography to aim the beam at a sample from various angles and employ the diffraction patterns to reconstruct the semiconductor's interior. Researcher Gabriel Aeppli says the technique yields better resolution than other methods, and at a larger scale. The conventional industry-standard technique involves progressively removing the processor's layers and capturing electron microscope images of one small area of the chip at a time. TechInsights' Dick James sees limitations to the Swiss team's work, including the need for a synchrotron source to generate the images. Still, the University of Southern California's Anthony Levi says, "total transparency in chip manufacturing is on the horizon. This is going to force a rethink of what computing is."
|
New Computer Software Program Excels at Lip Reading
University of Oxford
March 17, 2017
Researchers at the University of Oxford in the U.K. have developed Watch, Attend, and Spell (WAS), artificial intelligence (AI) software that uses computer vision and machine learning methods to understand how to lip read from a dataset of more than 5,000 hours of video footage containing more than 118,000 sentences and 17,500 words. The researchers compared the ability of WAS with a human lip-reading expert and found the software recognized 50 percent of the words in the dataset without error, compared to 12 percent for the human expert. They say WAS could be used to help hard-of-hearing individuals to navigate the world around them. "Lip reading is an impressive and challenging skill, so WAS can hopefully offer support to this task; for example, suggesting hypotheses for professional lip readers to verify using their expertise," says Oxford's Joon Son Chung.
|
Artificial Intelligence Is Learning to Predict and Prevent Suicide
Wired
Megan Molteni
March 17, 2017
Physicians at research hospitals and the U.S. Department of Veterans Affairs are testing suicide-prevention platforms driven by artificial intelligence (AI) to develop predictive models that can customize interventions earlier. Researchers at Florida State University trained algorithms on 2 million patients' anonymized digital health records so they can recognize which mix of variables best anticipate a suicide attempt. They say the machine learning AI can make such predictions with 80-percent to 90-percent accuracy as far off as two years in the future. Meanwhile, Cogito's Companion software app listens to a person's voice, collecting everything users say in a day, to find vocal indicators of depression and other mood shifts by analyzing the tone, fluidity of speaking, and levels of engagement with a conversation. "Having that kind of rich data is enormously powerful in understanding the nature of a mental health issue," notes David K. Ahern at Brigham and Women's Hospital.
|
Quantum Physics Offers Insights Into Music Expressivity
Queen Mary, University of London
March 15, 2017
Researchers at Queen Mary University of London (QMUL) in the U.K. have pioneered a new approach to understanding the musical experience by analyzing the vibrato effect using the Filter Diagonalization Method (FDM) in music signal processing. FDM was originally developed to investigate quantum dynamical resonances of atoms and molecules. "Although musical signals are very different from their quantum counterparts, mathematically they share many similarities, including the characteristics of their resonances," says QMUL's Khalid Rajab. He also notes the harmonics in musical signals can be more complex to analyze than their quantum equivalents. The researchers say FDM could help musicians realize the perfect vibrato, help sound artists produce more natural-sounding vibrato, and enable scientists to plot out stylistic trends in vibrato use across cultures and time. "We are now one step closer to understanding the mechanics of music communication, the nuances that performers introduce to the music, and the logic behind them," says QMUL professor Elaine Chew.
|
Computer Intelligence Spots Genes Linked to Duchenne Muscular Dystrophy
Business Standard
March 16, 2017
Researchers at the Maulana Abul Kalam Azad University of Technology in India say they used computational intelligence to identify a set of genes that are "collectively responsible" for Duchenne Muscular Dystrophy (DMD), a genetic disease causing progressive decay in muscles. The researchers mined a dataset of about 50,000 genes to obtain a subset containing 30 genes that are collectively discriminant in gene expression between DMD-affected patients and healthy patients. The researchers note the 30 identified genes could be clinically investigated to design new drugs or modify existing drugs for DMD. The team found the genes by developing two new algorithms based on existing meta-heuristic algorithms, which are designed to find good or near-optimal solutions at a reasonable computational cost and time. "Shortage/unavailability of data is always a limitation for computational research," says researcher Surama Biswas. "More methods and datasets may always be employed for validating the research."
|
Stanford Scientists Create Three-Dimensional Bladder Reconstruction
Stanford News
Jackie Flynn
March 16, 2017
Researchers at Stanford University have developed advanced computer-vision algorithms to three-dimensionally (3D) reconstruct bladders from endoscopic images, which could help physicians create organ maps, better prepare for surgical procedures, and detect early cancer recurrences. "The beauty of this project is that we can take data that doctors are already collecting," says Stanford professor Audrey Bowden. Her team employed video footage of a routine cytoscopy to obtain data their algorithms used to reconstruct a bladder's shape and internal semblance, and they say the method is potentially applicable to other hollow organs, such as the stomach or colon. The researchers tested the method's accuracy by first producing a model based on endoscopy images taken in a 3D-printed bladder, so they could directly compare their rendering to the actual organ. They determined the rendering's correspondence to the tissue phantom was mostly error-free.
|
Wi-Fi on Rays of Light: 100 Times Faster, and Never Overloaded
Eindhoven University of Technology (Netherlands)
March 17, 2017
Researchers at the Eindhoven University of Technology (TU/e) in the Netherlands have developed a Wi-Fi network based on directable infrared light rays, offering capacity topping 40 Gbps for each ray, and eliminating the need to share wavelengths since every device receives its own ray. Data is transmitted by a few centralized antennas equipped with passive diffraction gratings and optical fiber to direct the rays, so there are no moving parts and the system requires no power, according to TU/e's Joanne Oh. She says devices are assigned different wavelengths by the same antenna, which means there is no interference from a neighboring Wi-Fi network. Oh's work is part of TU/e's BROWSE project, which is focusing on the creation of indoor optical wireless networks. BROWSE differs from other research in this area with the invention of the diffraction gratings, which can simultaneously accommodate numerous light rays and devices.
|
Chemists Are First in Line for Quantum Computing's Benefits
Technology Review
Tom Simonite
March 17, 2017
Chemists are likely to be the first beneficiaries of practical quantum computers thanks to their ability to model molecules and reactions. "We have more confidence in the smaller systems for chemistry," says IBM's Scott Crowder. Meanwhile, Harvard University professor Alan Aspuru-Guzik says some potential applications of quantum computers include accelerating his group's research into finding new light-emitting molecules for displays, as well as enabling batteries for grid-scale energy storage. In addition, Microsoft is developing hybrid quantum/conventional computers to simulate chemical processes, with one potential area of concentration being new superconducting materials, says Microsoft's Krysta Svore. Modeling the quantum effects that influence molecular structures and reactions comes naturally to quantum computers, because they can encode data into those same quantum states. The computers' constituent qubits can exploit quantum-mechanical processes to take computational shortcuts that conventional computers cannot support.
|
A Scientist and a Supercomputer Re-create a Tornado
UW-Madison News
Eric Verbeten
March 13, 2017
Researchers at the University of Wisconsin-Madison are using supercomputer simulations to study the structure of tornado-producing supercell thunderstorms. The researchers say they can create in-depth visualizations of supercells and discern how they form and ultimately spawn tornadoes. The most recent simulation recreates the "El Reno" tornado, which touched down in Oklahoma in 2011 and caused damage over a 63-mile area. Using real-world observational data, the researchers were able to recreate the weather present at the time of the storm and witness the steps leading up to the creation of the tornado. The simulation reveals in high resolution the numerous "mini-tornadoes" that form at the onset of the main tornado. As the funnel cloud develops, the mini-tornadoes begin to merge, adding strength to the storm. The researchers used the Blue Waters Supercomputer housed at the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign.
|
Computing With Spiders' Webs
University of Bristol News
March 13, 2017
Researchers at the universities of Bristol and Oxford in the U.K. are examining spiders' webs to determine their computational capabilities, and based on this research they will develop new sensor technology to measure vibrations and flow. The researchers will examine the structure of spiders' webs to understand how their designs could be used as a computer, a line of research known as morphological computation. Morphological computation is a design approach, often used in robotics, that considers the body of a robot vital for any intelligent behavior. "The idea of intelligent morphological structures is not just useful for spiders, but can be developed into novel, intelligent sensor technologies, especially for vibration and flow sensors," says Bristol's Helmut Hauser. The researchers believe the results of the study could provide insights into the way morphological features of biological sensors are utilized in bio-inspired sensor design.
|
|
