A team of leading computer scientists has developed a novel computational system to aid the design and fabrication of wind-up toys, focusing on automating the intricate interior machinery responsible for the toys' wind-up motion.
Rice University materials scientists lead a project to turn strong, light and compressible schwarzites from theory to reality with three-dimensional printers. The resulting materials share their properties from the nano- to the macroscale.
Can a close look at the universe give us solutions to problems too difficult for a computer -- even if we built a computer larger than a planet? Physicist Stephen Jordan reflects on this question in a new NIST video, along with a scientific paper that considers one particular tough problem the universe might answer.
Since the late 60's electronic devices have stored and transmitted information (bits) in two-dimensional circuits. Now, researchers at the University of Cambridge have been able to break this barrier by creating a nanoscale magnetic circuit capable of moving information along the three dimensions of space. This breakthrough could lead to an important increase in storage and processing capacities of electronic devices over those used today.
In 'A Theory of How Columns in the Neocortex Enable Learning the Structure of the World,' published in Frontiers in Neural Circuits, Numenta researchers propose a new theory for how the brain learns models of objects through movement. The paper introduces a missing ingredient to sensory input processing: a location signal relative to the object being sensed. An accompanying repository containing the paper's source code, data and scripts allows other scientists to reproduce the results.
UW researchers who used machine learning tools to analyze language in 800 Hollywood movie scripts found subtle but widespread gender bias in the amount of power and agency given to male and female characters.
Researchers from UC Berkeley, UC Davis and TACC used Stampede2 to complete a 100-epoch ImageNet deep neural network training in 11 minutes -- the fastest time recorded to date. Using 1600 Skylake processors they also bested Facebook's prior results by finishing a 90-epoch ImageNet training with ResNet-50 in 32 minutes. Given TACC's large user base and huge capacity, this capability will have a major impact across all fields of science.
An economy based on zero growth could be more stable -- experiencing fewer crashes -- and bring higher wages, suggests a new University of Sussex study.
Human traits such as the face, voice, and gait are commonly used for biometric-based person recognition. Although gait is one of the most practical traits for video-based surveillance and forensics, this approach is susceptible to intra-subject variations. Osaka University-based researchers proposed a new convolutional neural network-based approach to gait recognition, which exhibits robustness (insensitiveness) to spatial displacement. Their approach outperformed current benchmarks in terms of verification/identification tasks and view differences.
An inexpensive device could finally solve the problem of improving wireless signal strength and security for indoor spaces with multiple rooms.