The ImageNet Large Scale Visual RecognitionChallenge is a benchmark in object category classification and detection on hundreds of object categories and millions of images. The challenge has been run annually from 2010 to present, attracting participation from more than fifty institutions. This paper describes the creation of this benchmark dataset and the advances in object recognition that have been possible as a result. We discuss the challenges of collecting large-scale ground truth annotation, highlight key breakthroughs in categorical object recognition, provide a detailed analysis of the current state of the field of large-scale image classification and object detection, and compare the state-ofthe-art computer vision accuracy with human accuracy. We conclude with lessons learned in the 5 years of the challenge, and propose future directions and improvements.

No abstract

Low-cost, high-throughput gene synthesis and precise control of protein expression are of critical importance to synthetic biology and biotechnology. Here we describe the development of an on-chip gene synthesis technology, which integrates on a single microchip the synthesis of DNA oligonucleotides using inkjet printing, isothermal oligonucleotide amplification and parallel gene assembly. Use of a mismatch-specific endonuclease for error correction results in an error rate of ~0.19 errors per kb. We applied this approach to synthesize pools of thousands of codon-usage variants of lacZα and 74 challenging Drosophila protein antigens, which were then screened for expression in Escherichia coli. In one round of synthesis and screening, we obtained DNA sequences that were expressed at a wide range of levels, from zero to almost 60% of the total cell protein mass. This technology may facilitate systematic investigation of the molecular mechanisms of protein translation and the design, construction and evolution of macromolecular machines, metabolic networks and synthetic cells.

Despite theoretical predictions, it remains an experimental challenge to realize an artificial magnetic skyrmion whose topology can be well controlled and tailored so that its topological effect can be revealed explicitly in a deformation of the spin textures. Here we report epitaxial magnetic thin films in which an artificial skyrmion is created by embedding a magnetic vortex into an out-of-plane aligned spin environment. By changing the relative orientation between the central vortex core polarity and the surrounding out-of-plane spins, we are able to control and tailor the system between two skyrmion topological states. An in-plane magnetic field is used to annihilate the skyrmion core by converting the central vortex state into a single domain state. Our result shows distinct annihilation behaviour of the skyrmion core for the two different skyrmion states, suggesting a topological effect of the magnetic skyrmions in the core annihilation process.

The past couple of years saw exciting new developments in microchip-based gene synthesis technologies. Such technologies hold the potential for significantly increasing the throughput and decreasing the cost of gene synthesis. Together with more efficient enzymatic error correction and genome assembly methods, these new technologies are pushing the field of synthetic biology to a higher level.

We examine the effectiveness of opportunistic use of reverse body bias (RBB) to reduce leakage power during active operation, burn-in, and standby in 0.18µm single-V t and 0.13µm dual-V t logic process technologies. We investigate its dependencies on channel length, target V t , temperature and technology generation. We show that RBB becomes less effective for leakage reduction at shorter channel lengths and lower V t at both high and room temperatures, especially when target intrinsic leakage currents are high. RBB effectiveness also diminishes with technology scaling primarily because of worsening short-channel effects (SCE), particularly when target V t values are low. We present a model that relates different transistor leakage components to full-chip leakage current, and validate the model through testchip measurements across a range of RBB values.

Accurate, economical and high-throughput gene and genome synthesis is essential to the development of synthetic biology and biotechnology. New large scale gene synthesis methods harnessing the power of DNA microchips have recently been demonstrated. Yet, the technology is still compromised by a high occurrence of errors in the synthesized products. These errors still require substantial effort to correct. To solve this bottleneck, novel approaches based on new chemistry, enzymology or next generation sequencing have emerged. This review discusses these new trends and promising strategies of error-filtration, error correction and error-prevention in de novo gene and genome synthesis. Continued innovation in error correction technologies will enable affordable and large scale gene and genome synthesis in the near future.

No abstract