10 Bits: the Data News Hotlist
This week’s list of data news highlights covers June 3-9, 2017, and includes articles about researchers using analytics to study Sudden Infant Death Syndrome and Malaysia’s new commitments to open data.
Duke University mathematics professor Jonathan Mattingly has developed an algorithm that can generate alternative state election maps designed to measure and highlight the distortive effect gerrymandering—the redrawing of voting districts to benefit certain candidates or parties—has on election outcomes in the United States. Though gerrymandering is common, it can be difficult to curb because there are not standardized methods for drawing voting districts. Mattingly will testify as an expert witness in North Carolina this summer in a case involving racially-biased gerrymandering, which could pave the way for a more data-driven approach to drawing voting maps.
An international team of researchers led by Imperial College London have developed a machine learning technique to predict the properties of x-rays used in advanced measuring techniques, enabling faster and more accurate data collection. Scientists use free electron lasers (FELs) to target pulses of x-rays at molecular structures to measure their physical and chemical properties, however FELs can be unstable and creating unusable data. Using machine learning, the researchers were able to predict certain properties of x-rays using initial measurements and ensure the data is accurate at a much higher rate, enabling them to take advantage of the next generation of FELs that can generate hundreds of thousands of x-ray pulses per second, compared to the current generation which use just 100 pulses per second.
Facebook has developed a machine learning image recognition system that can analyze 40,000 images per second to train itself. Though Facebook has plenty of image data at its disposal to train AI systems, hardware limitations can make analyzing large datasets very time intensive. Facebook was able to achieve this speed by designing a machine learning model that can take advantage of its new graphical processing unit (GPU) hardware stack named Big Basin, the designs for which which it plans to make available as open source.
Researchers at Carnegie Mellon University and Google have developed a system that combines machine learning and game theory to pit robotic arms against each other to accelerate the process of learning to grasp objects. Robotic grasping systems often try and fail thousands of times to learn how to properly grasp objects. The researchers instead used two robotic arms—one programmed to grasp an object and one trying to disrupt the former—to provide stronger signals to a neural network about what a successful grasp is. With this adversarial approach, the researchers were able to achieve a grasp success rate of 52 percent after 6,000 attempts, compared to a traditional approach resulting in 47 percent grasp rate with 16,000 tries.
Microsoft has partnered with Seattle Children’s Hospital to develop and deploy data science tools that can help medical professionals better understand and prevent Sudden Infant Death Syndrome (SIDS). Because SIDS is a combination of a variety of different factors, it can be difficult to identify what puts infants at risk through traditional methods. Researchers at Microsoft and the hospital will use machine learning to analyze 29 million records from the U.S. Centers for Disease Control and Prevention to pinpoint these risk factors, which smaller-scale medical studies would be unlikely to uncover, and identify potential preventive measures, such as the ideal number of prenatal doctor visits for expecting mothers.
The Malaysian government has completed the World Bank’s Open Data Readiness Assessment, which assesses the readiness of governments to publish open data to improve governance and spur economic growth, and it has committed to implementing the assessment’s recommendations. The assessment found the Malaysian government met the necessary criteria to expand its open data initiatives, including funding, demand for open data, and leadership, and recommended that Malaysia direct all agencies to inventory their data assets and treat data as open by default. Malaysia participated in the assessment to help it achieve its goal of moving from 51st place to the top 30 by 2020 in the Global Open Data Barometer, a ranking of countries’ open data policies and practices.
Washington Governor Jay Inslee has signed an executive order to permit the testing of self-driving cars without humans in the driver seat within two months. While most jurisdictions allow for the testing of autonomous vehicles, they typically require a human driver to take over control in the event of an emergency. Inslee’s order will allow autonomous vehicle companies to apply for permits to pilot self-driving cars without a human operator, provided the vehicle meets certain safety standards.
Facebook has announced it will share mapping data with emergency responders after a disaster to help fill information gaps. Facebook will create disaster maps that use aggregated, de-identified data from Facebook users that can convey population density of a specific area, the flow of people, and the number of people who have checked-in as safe, which can be crucial information in the wake of a disaster and can help responders better prioritize their efforts. Facebook will initially share this data with UNICEF, the International Federation of the Red Cross and Red Crescent Societies, and the World Food Programme.
The U.S. National Institutes of Health (NIH) has launched the beta phase of the “All of Us” research program, which aims to build a million-person research cohort to support its Precision Medicine Initiative. Over the next several months, NIH will recruit up to 15,000 people throughout the country to enroll in the program so NIH can identify problems in the process and improve them for the full launch, which it expects to start in late 2017 or early 2018.
A nonprofit organization called Front Porch that manages an elderly community is piloting a program that provides its elderly residents with Amazon’s Echo smart home device and its accompanying AI assistant Alexa to help them stay better connected to the world. Because many of Front Porch’s residents have physical ailments such as macular degeneration or hand tremors that make it difficult for them to interact with technology through keyboards and touchscreens, Alexa can help users communicate and access online services. Front Porch will soon expand the pilot to connect Alexa to other smart home devices, including smart thermostats and outlets.
Image: China Crisis.