'Virtual Safe Space' to Help Bumblebees

University of Exeter

A "virtual safe space" developed at the University of Exeter in the U.K. tests the threats with which bumblebees are confronted, creating a computer model of bumblebee population dynamics in response to variables such as pesticides, parasites, and loss of habitat. The predictive Bumble-BEEHAVE tool is described by Exeter's Grace Twiston-Davies as "a free, user-friendly system" that “takes into account the many complicated factors that interact to affect bumblebees. This provides a virtual safe space to test the different management options." Bumble-BEEHAVE can model the growth, behavior, and survival of six British bumblebee species in an environment that has various nectar and pollen sources. Exeter's Juliet Osborne notes the simulation "enables researchers to understand the individual and interacting effects of the multiple stressors affecting bumblebee survival and the feedback mechanisms that may buffer a colony against environmental stress, or indeed lead to spiraling colony collapse." Twiston-Davies says the researchers are already working with land managers and wildlife groups on the ground.

From "'Virtual Safe Space' to Help Bumblebees"
University of Exeter (05/23/18)
View Full Article

How to Evaluate Machine Learning? U of T Research Supports Latest Benchmark Initiative

U of T News

An industrial-academic consortium that includes Google, the University of Toronto (U of T) in Canada, and Harvard and Stanford universities is developing a new benchmark suite for assessing machine learning (ML) performance. U of T's Gennady Pekhimenko says the MLPerf consortium is investigating two benchmarking areas--an "open" category in which any model can be applied to a fixed dataset, and a "closed" category in which both model and datasets are fixed, making execution time, power requirements, and design-cost evaluations helpful. Pekhimenko notes his laboratory has developed an open source benchmark suite called TBD (To Be Determined) as a training benchmark for deep neural networks. "We're interested in understanding how well available hardware and software perform, but we also look at both hardware and software efficiency," he says. "We then provide hints to the ML developers, so they can make their networks more efficient, and hence develop new algorithms and insights faster."

From "How to Evaluate Machine Learning? U of T Research Supports Latest Benchmark Initiative"
U of T News (05/22/18) Nina Haikara
View Full Article

Academy to Deliver New Initiatives for Women and Girls in STEM

Australian Academy of Science

The Australian government has allocated $600,000 to the Australian Academy of Science to develop a 10-year roadmap for sustained increases in engagement and participation of girls and women in science, technology, engineering, and math (STEM). "The barriers to equity in STEM are compounded by the gendered portrayal of science in the media, and a preponderance of male scientists on STEM committees and boards across government, academia, and industry," says Academy president Andrew Holmes. The new funding, along with a $250,000 Women in STEM and Entrepreneurship grant, will enable the Academy to deliver a national #WomenInSTEMOnline online portal for female STEM professionals. The #WomenInSTEMOnline project will promote gender equity by improving women's participation in STEM conferences and panels, science and technology media, government and industry committees, boards, and STEM award nominations and applications.

From "Academy to Deliver New Initiatives for Women and Girls in STEM"
Australian Academy of Science (05/21/18)
View Full Article

Tunable Third Harmonic Generation in Graphene Paves the Way to High-Speed Optical Communications and Signal Processing

Graphene Flagship

For the first time, researchers have demonstrated gate tunable third harmonic generation (THG) in graphene. Conducted by researchers at Graphene Flagship partner University of Cambridge, working with researchers at the Italian universities Politecnico di Milano, and Istituto Italiano di Tecnologia in Genoa, this work could lead to on-chip broadband optical switches for data transport in optical systems. The researchers proved the strong THG in graphene can be controlled by an external electric field, and that ultra-broad bandwidth can boost efficiency. This type of THG optical switch could offer new insights into matter by making more "colors" available for use in spectroscopy. In addition, graphene THG optical switches could leverage previously unused optical frequencies to transmit data along optical cables, increasing the amount of data that can be transmitted and raising data speeds. As a next step, the researchers say they want to move closer to producing integrated devices in optical fibers and waveguides.

From "Tunable Third Harmonic Generation in Graphene Paves the Way to High-Speed Optical Communications and Signal Processing"
Graphene Flagship (05/21/18) Sian Fogden
View Full Article

Computer Redesigns Enzyme

University of Groningen

Using a computational method, biotechnologists at the University of Groningen in the Netherlands redesigned aspartase and converted it to a catalyst for asymmetric hydroamination reactions. Fellow researchers at the Chinese Academy of Sciences scaled up the enzyme's production and created kilograms of pure building blocks for pharmaceuticals and other bioactive compounds. Modifying enzymes is labor-intensive and can require testing thousands of enzyme variants in multiple rounds. A more efficient approach would be to make a rational design of the required changes based on information on the enzyme's structure and properties, but even this leads to a huge matrix of protein structures, explains University of Groningen's Hein Wijma. To speed up this process, the researchers used an extremely fast Monte Carlo search algorithm that seeks trends in the enzyme's reactivity. Performing the search took a couple of days in a dedicated computer cluster at the university. The method achieved substrate conversions of 99% with a 99% enantioselectivity in quantities up to one kilogram, proving that the method works for producing useful enzymes.

From "Computer Redesigns Enzyme"
University of Groningen (05/21/18)
View Full Article

The Way Toddlers Waddle Can Teach Robot Footballers How to Play

New Scientist

A team of researchers at New York University and the University of Texas at Austin has determined that a team of robot football (soccer) players trained to mimic how infant humans walk comprehensively can beat other robots trained on geometric walking patterns. The researchers recorded the walking paths of 90 babies at play and used the data to create an hour-long training course in a three-dimensional simulator. Each humanoid robot was rewarded or penalized based on how well they followed the course. A team of robots trained to walk similarly to infants then played teams trained on simpler, straight, square, and circular courses in a simulation of the RoboCup. Each team played every other team 1,000 times, with the infant-trained team finishing with 2,888 wins, 1,037 ties, and only 75 losses.

From "The Way Toddlers Waddle Can Teach Robot Footballers How to Play"
New Scientist (05/18/18) Edd Gent
View Full Article - May Require Paid Subscription

Keep the Light Off: A Material With Improved Mechanical Performance in the Dark

Nagoya University (Japan)

Researchers at Nagoya University in Japan have discovered that an inorganic zinc sulfide (ZnS) semiconductor behaves differently in the dark than in the light. The researchers found ZnS crystals were brittle when exposed to light, but flexible when kept in the dark at room temperature. The ZnS crystals that were kept in the dark deformed plastically without fracturing, up to a strain of 45%. The teams attributed the increased plasticity of the crystals in the dark to the high mobility of dislocations in complete darkness. The ZnS crystals kept in the light were brittle because their deformation mechanism was different from those in the dark. In addition, the ZnS crystals in the dark exhibited a significant decrease in the band gap of the deformed crystals, meaning their electrical conductivity can be controlled by mechanical deformation in the dark.

From "Keep the Light Off: A Material With Improved Mechanical Performance in the Dark"
Nagoya University (Japan) (05/18/18)
View Full Article

The Percentage of Open Source Code in Proprietary Apps Is Rising

Help Net Security

Proprietary software applications exhibit an increasing volume of open source code, along with the elevated danger of those apps being hacked, according to a new study. Black Duck On-Demand's 2017 audit of the anonymized data of more 1,100 commercial codebases revealed that 96% of those apps had open source components, while the average percentage of open source in those codebases increased since last year from 36% to 57%, implying that many apps now contain significantly more open source than proprietary code. Analysts at the Synopsys Center for Open Source Research & Innovation cite lower development costs, accelerated time to market, and expedited developer productivity as reasons for pervasive open source code use. They found at least one vulnerability in 78% of the codebases studied, with an average of 64 vulnerabilities for every codebase; an average 77% of the codebase for Internet of Things apps contained open source code, with an average 677 vulnerabilities per app.

From "The Percentage of Open Source Code in Proprietary Apps Is Rising"
Help Net Security (05/22/18) Zeljka Zorz
View Full Article

NC State Becomes First Quantum Computing Hub in North America

Technician Online (NC)

North Carolina State University (NC State) has been chosen by IBM as the first North American institution to become a quantum computing hub (Q Hub) as part of IBM's Q Network. Through the Q Hub, students and faculty can access IBM's 20-quantum-bit supercomputer to solve equations that classical systems would take too long to compute. "No way can we create a computer with 10% of all the atoms of the earth," notes IBM Research's Bob Sutor. "Yet, with a quantum computer...we can represent all that information exactly." IBM plans to work with Fortune 500 companies and universities to address larger problems more rapidly via the Q Network, while Sutor sees NC State collaborating with local organizations to solve complex equations. NC State's faculty is developing courses that both undergraduates and graduates can use to build their quantum-computing knowledge in anticipation of the hub's operational debut this fall.

From "NC State Becomes First Quantum Computing Hub in North America"
Technician Online (NC) (05/20/18) Alicia Thomas
View Full Article

How AI and IoT Smart Sensors Are Preserving a Lake Ecosystem

Robotics Business Review

IBM Research, Rensselaer Polytechnic Institute, and The Jefferson Project have been watching Lake George, NY, and its ecosystem for five years as part of a three-decade study of the lake's water quality and other environmental forces. The partners have deployed a sensor network composed of platforms, weather stations, tributary monitors, and acoustic Doppler current profiles to supply information at more than 40 sites. This network constantly measures weather, water movement, and water quality, sending that data to an IBM Research facility that generates high-resolution computational modeling and forecasts. The sensors apply these findings to respond to the physical data and models to accrue more data or notify project members of anomalies. The researchers believe the intelligence culled from the sensor network can yield better insights on processes and events that could impact the Lake George ecosystem.

From "How AI and IoT Smart Sensors Are Preserving a Lake Ecosystem"
Robotics Business Review (05/15/18) Keith Shaw
View Full Article

Princeton Researchers Crowdsource Brain Mapping With Gamers, Discover Six New Neuron Types

Princeton University

Princeton University researchers have generated and shared maps of approximately 1,000 neurons using data crowdsourced by more than 265,000 online video-game players, with Princeton's Sebastian Seung noting the team has classified between 35 and 50 types of ganglion cells. The Eyewire Museum is an interactive archive of neurons that the general public and neuroscientists worldwide can access. The Eyewire gamer community developed the neural maps, using data from a mouse retina they traced with machine learning. Since 2012, participants have colored in more than 10 million "cubes," and plotted more than 3,000 neurons; each cell is vetted by five to 25 gamers before it is deemed completed and added to the museum. Princeton's Shang Mu says six of the ganglions classified by Eyewire appear to be unique, in that no matching descriptions turned up in a literature search.

From "Princeton Researchers Crowdsource Brain Mapping With Gamers, Discover Six New Neuron Types"
Princeton University (05/17/18) Liz Fuller-Wright
View Full Article

What's Trending in Fake News? IU Tools Show Which Stories Go Viral, and If 'Bots' Are to Blame

Indiana University

Indiana University (IU) researchers recently launched upgrades to the Hoaxy and Botometer online tools, as well as a new educational game, to make people more responsible as news consumers. "The majority of the changes to Hoaxy and Botometer are specifically designed to make the tools more usable by journalists and average citizens," says IU's Filippo Menczer. Hoaxy, a search engine that shows users how stories from low-credibility sources spread on Twitter, now shows users which stories are trending on the social media site and signals what proportion of users spreading the stories are likely to be bots. Botometer, which assigns a score to Twitter users based on the probability that the account is automated, now has updated machine learning algorithms to identify bots with greater accuracy. Meanwhile, IU's "Fakey" literacy game combines news stories with falsehoods, giving players points for fact-checking bogus information and liking or sharing accurate stories.

From "What's Trending in Fake News? IU Tools Show Which Stories Go Viral, and If 'Bots' Are to Blame"
Indiana University (05/17/18) Kevin Fryling
View Full Article

UCLA Engineer Develops 3D Printer That Can Create Complex Biological Tissues

UCLA Samueli Newsroom

Ali Khademhosseini of the University of California, Los Angeles led a study describing a new three-dimensional (3D) printing method to construct therapeutic biomaterials from multiple materials. His team employed stereolithography in conjunction with a customized 3D printer Khademhosseini designed, which uses a custom-built microfluidic chip with inlets that each produce a distinct material, combined with a digital micromirror array. The team used varying hydrogels that cohere into scaffolds for tissue, while the micromirrors steered light onto the printing surface to mark the outline of the object. Illumination also catalyzed the formation of molecular bonds in the materials, inducing hydrogel solidification. The mirror array re-directed the light pattern during printing to indicate the contours of each new layer. The process was initially used to produce simple shapes, and then applied to complex 3D structures to emulate muscle tissue and muscle-skeleton connective tissues.

From "UCLA Engineer Develops 3D Printer That Can Create Complex Biological Tissues"
UCLA Samueli Newsroom (05/17/18)
View Full Article

Google's DeepMind Is Using Neural Nets to Explore Dopamine's Role in Learning

VentureBeat

In an effort to understand how people learn rapidly from example through a process called meta-learning, researchers at Google subsidiary DeepMind in London modeled human physiology using a recurrent neural network, which internalizes past actions and observations and draws from those experiences. The network's reward prediction error, the signal that mathematically optimizes the algorithm over time through trial and error, acted as dopamine, the brain chemical believed to play a key role in the learning process. The researchers compared the system's performance to that of animals in six neuroscientific meta-learning experiments. One of the tests involved choosing between two randomly selected images, one of which was associated with a reward. Like the animal subjects, the algorithm learned to make reward-associated choices. Learning occurred in the recurrent neural network, supporting the theory that dopamine is critical to meta-learning. The consistency of the neural network's behavior suggests that dopamine also conveys and encodes information about tasks and rule structures. The work highlights the potential of neural network research in the field of medicine.

From "Google's DeepMind Is Using Neural Nets to Explore Dopamine's Role in Learning"
VentureBeat (05/14/18) Kyle Wiggers
View Full Article