Featured
-
-
Article
| Open AccessAvoiding fusion plasma tearing instability with deep reinforcement learning
Artificial intelligence control is used to avoid the emergence of disruptive tearing instabilities in the magnetically confined fusion plasma in the DIII-D tokamak reactor.
- Jaemin Seo
- , SangKyeun Kim
- & Egemen Kolemen
-
Article
| Open AccessSolving olympiad geometry without human demonstrations
A new neuro-symbolic theorem prover for Euclidean plane geometry trained from scratch on millions of synthesized theorems and proofs outperforms the previous best method and reaches the performance of an olympiad gold medallist.
- Trieu H. Trinh
- , Yuhuai Wu
- & Thang Luong
-
Article |
Discovery of a structural class of antibiotics with explainable deep learning
An explainable deep learning model using a chemical substructure-based approach for the exploration of chemical compound libraries identified structural classes of compounds with antibiotic activity and low toxicity.
- Felix Wong
- , Erica J. Zheng
- & James J. Collins
-
Article
| Open AccessHuman mobility networks reveal increased segregation in large cities
There is extreme socioeconomic segregation in large US cities, arising from a greater choice of differentiated spaces targeted to specific socioeconomic groups, which can be countered by positioning city hubs (such as shopping centres) to bridge diverse neighbourhoods.
- Hamed Nilforoshan
- , Wenli Looi
- & Jure Leskovec
-
Article
| Open AccessAn analog-AI chip for energy-efficient speech recognition and transcription
A low-power chip that runs AI models using analog rather than digital computation shows comparable accuracy on speech-recognition tasks but is more than 14 times as energy efficient.
- S. Ambrogio
- , P. Narayanan
- & G. W. Burr
-
Article
| Open AccessSkilful nowcasting of extreme precipitation with NowcastNet
A new nowcasting model unifies physical-evolution schemes and deep-learning methods to accurately predict precipitation with lead times of up to 3 h, including extreme-precipitation events and weather systems that were previously considered intractable with physics-based numerical methods.
- Yuchen Zhang
- , Mingsheng Long
- & Jianmin Wang
-
Article
| Open AccessHealth system-scale language models are all-purpose prediction engines
A clinical language model trained on unstructured clinical notes from the electronic health record enhances prediction of clinical and operational events.
- Lavender Yao Jiang
- , Xujin Chris Liu
- & Eric Karl Oermann
-
Article |
Aerial additive manufacturing with multiple autonomous robots
An additive manufacturing method using a team of autonomous aerial robots allows for scalable and adaptable three-dimensional printing, and is used to deposit building materials during flight.
- Ketao Zhang
- , Pisak Chermprayong
- & Mirko Kovac
-
Article
| Open AccessDeep physical neural networks trained with backpropagation
A hybrid algorithm that applies backpropagation is used to train layers of controllable physical systems to carry out calculations like deep neural networks, but accounting for real-world noise and imperfections.
- Logan G. Wright
- , Tatsuhiro Onodera
- & Peter L. McMahon
-
Article
| Open AccessGlobal potential for harvesting drinking water from air using solar energy
Mapping of the global potential of atmospheric water harvesting using solar energy shows that it could provide safely managed drinking water for a billion people worldwide based on climate suitability.
- Jackson Lord
- , Ashley Thomas
- & Philipp H. Schmaelzle
-
Article |
A graph placement methodology for fast chip design
Machine learning tools are used to greatly accelerate chip layout design, by posing chip floorplanning as a reinforcement learning problem and using neural networks to generate high-performance chip layouts.
- Azalia Mirhoseini
- , Anna Goldie
- & Jeff Dean
-
Article |
AI-based pathology predicts origins for cancers of unknown primary
A deep-learning-based algorithm uses routinely acquired histology slides to provide a differential diagnosis for the origin of the primary tumour for complicated cases of metastatic tumours and cancers of unknown primary origin.
- Ming Y. Lu
- , Tiffany Y. Chen
- & Faisal Mahmood
-
Article |
A quantum magnetic analogue to the critical point of water
The pressure dependence and magnetic field dependence of the specific heat of a quantum magnet, SrCu2(BO3)2, demonstrate that its phase diagram contains a line of first-order transitions terminating at a critical point, in analogy with water.
- J. Larrea Jiménez
- , S. P. G. Crone
- & F. Mila
-
Article |
An autonomous debating system
An artificial intelligence system that can engage in a competitive debate with humans is presented.
- Noam Slonim
- , Yonatan Bilu
- & Ranit Aharonov
-
Article |
Towards real-time photorealistic 3D holography with deep neural networks
A deep-learning-based approach using a convolutional neural network is used to synthesize photorealistic colour three-dimensional holograms from a single RGB-depth image in real time, and termed tensor holography.
- Liang Shi
- , Beichen Li
- & Wojciech Matusik
-
Article |
First return, then explore
A reinforcement learning algorithm that explicitly remembers promising states and returns to them as a basis for further exploration solves all as-yet-unsolved Atari games and out-performs previous algorithms on Montezuma’s Revenge and Pitfall.
- Adrien Ecoffet
- , Joost Huizinga
- & Jeff Clune
-
Article |
Generating conjectures on fundamental constants with the Ramanujan Machine
An approach called the Ramanujan Machine demonstrates the use of algorithms to find mathematical conjectures in the form of formulas of fundamental constants, some of which remain unproved.
- Gal Raayoni
- , Shahar Gottlieb
- & Ido Kaminer
-
Matters Arising |
The effect of interventions on COVID-19
- Kristian Soltesz
- , Fredrik Gustafsson
- & Bo Bernhardsson
-
Article |
Mastering Atari, Go, chess and shogi by planning with a learned model
A reinforcement-learning algorithm that combines a tree-based search with a learned model achieves superhuman performance in high-performance planning and visually complex domains, without any knowledge of their underlying dynamics.
- Julian Schrittwieser
- , Ioannis Antonoglou
- & David Silver
-
Article |
Mobility network models of COVID-19 explain inequities and inform reopening
An epidemiological model that integrates fine-grained mobility networks illuminates mobility-related mechanisms that contribute to higher infection rates among disadvantaged socioeconomic and racial groups, and finds that restricting maximum occupancy at locations is especially effective for curbing infections.
- Serina Chang
- , Emma Pierson
- & Jure Leskovec
-
Review Article
| Open AccessArray programming with NumPy
NumPy is the primary array programming library for Python; here its fundamental concepts are reviewed and its evolution into a flexible interoperability layer between increasingly specialized computational libraries is discussed.
- Charles R. Harris
- , K. Jarrod Millman
- & Travis E. Oliphant
-
Letter |
Deep learning for multi-year ENSO forecasts
A statistical forecast model using a deep-learning approach produces useful forecasts of El Niño/Southern Oscillation events with lead times of up to one and a half years.
- Yoo-Geun Ham
- , Jeong-Hwan Kim
- & Jing-Jia Luo
-
Letter |
The seventh inner moon of Neptune
Hubble Space Telescope observations of the seventh inner moon of Neptune, Hippocamp, show that it is smaller than the other six, orbits near Proteus and probably originates from a fragment of Proteus.
- M. R. Showalter
- , I. de Pater
- & R. S. French
-
Letter |
A structural transition in physical networks
A modelling framework is presented to determine the optimal layout and physical properties of networks in which the nodes and links have physical sizes and intersections between components is prohibited.
- Nima Dehmamy
- , Soodabeh Milanlouei
- & Albert-László Barabási
-
Letter |
Glider soaring via reinforcement learning in the field
A reinforcement learning approach allows a suitably equipped glider to navigate thermal plumes autonomously in an open field.
- Gautam Reddy
- , Jerome Wong-Ng
- & Massimo Vergassola
-
Letter |
A per-cent-level determination of the nucleon axial coupling from quantum chromodynamics
Lattice quantum chromodynamics and a method inspired by the Feynman–Hellmann theorem are used to make a theoretical determination of the nucleon axial coupling with a precision of one per cent, giving the value 1.271 ± 0.013.
- C. C. Chang
- , A. N. Nicholson
- & A. Walker-Loud
-
Letter |
Image reconstruction by domain-transform manifold learning
Image reconstruction is reformulated using a data-driven, supervised machine learning framework that allows a mapping between sensor and image domains to emerge from even noisy and undersampled data, improving accuracy and reducing image artefacts.
- Bo Zhu
- , Jeremiah Z. Liu
- & Matthew S. Rosen
-
Letter |
Confocal non-line-of-sight imaging based on the light-cone transform
A confocal scanning technique solves the reconstruction problem of non-line-of-sight imaging to give fast and high-quality reconstructions of hidden objects.
- Matthew O’Toole
- , David B. Lindell
- & Gordon Wetzstein
-
Letter |
In situ X-ray diffraction measurement of shock-wave-driven twinning and lattice dynamics
In situ femtosecond X-ray diffraction measurements reveal that the dominant mechanism of shock-wave-driven deformation in tantalum changes from twinning to dislocation slip as pressure increases.
- C. E. Wehrenberg
- , D. McGonegle
- & J. S. Wark
-
Article |
Mastering the game of Go without human knowledge
Starting from zero knowledge and without human data, AlphaGo Zero was able to teach itself to play Go and to develop novel strategies that provide new insights into the oldest of games.
- David Silver
- , Julian Schrittwieser
- & Demis Hassabis
-
Letter |
Solving a Higgs optimization problem with quantum annealing for machine learning
A machine learning algorithm implemented on a quantum annealer—a D-Wave machine with 1,098 superconducting qubits—is used to identify Higgs-boson decays from background standard-model processes.
- Alex Mott
- , Joshua Job
- & Maria Spiropulu
-
Letter |
Magnetic reversals from planetary dynamo waves
Polarity reversals caused by dynamo waves are demonstrated in a magnetohydrodynamic model that is relevant to planetary cores, suggesting a possible mechanism of geomagnetic reversals.
- Andrey Sheyko
- , Christopher C. Finlay
- & Andrew Jackson
-
Letter |
Dynamics from noisy data with extreme timing uncertainty
A data-analytical approach that can extract the history and dynamics of complex systems from noisy snapshots on timescales much shorter than the uncertainty with which the data were recorded is described; the approach is demonstrated by extracting the dynamics on the few-femtosecond timescale from experimental data recorded with 300-femtosecond timing uncertainty.
- R. Fung
- , A. M. Hanna
- & A. Ourmazd
-
Letter |
Exploring the quantum speed limit with computer games
The crowd sourcing and gamification of a problem in quantum computing are described; human players succeed in solving the problem where purely numerical optimization fails, providing insight into, and a starting point for, strategies for optimization.
- Jens Jakob W. H. Sørensen
- , Mads Kock Pedersen
- & Jacob F. Sherson
-
Article |
Mastering the game of Go with deep neural networks and tree search
A computer Go program based on deep neural networks defeats a human professional player to achieve one of the grand challenges of artificial intelligence.
- David Silver
- , Aja Huang
- & Demis Hassabis
-
Letter |
Ab initio alpha–alpha scattering
An ab initio calculation of alpha–alpha scattering is described for which the number of computational operations scales approximately quadratically with particle number and which uses lattice Monte Carlo simulations and lattice effective field theory, combined with the adiabatic projection method to reduce the eight-body system to a two-cluster system.
- Serdar Elhatisari
- , Dean Lee
- & Ulf-G. Meißner
-
Letter |
Hybrid shallow on-axis and deep off-axis hydrothermal circulation at fast-spreading ridges
High-resolution three-dimensional simulations of hydrothermal flow beneath fast-spreading ridges predict two interacting flow components — shallow on-axis flow and deeper off-axis flow — that merge to feed axial vent sites, reconciling previously incompatible models favouring only one flow component.
- Jörg Hasenclever
- , Sonja Theissen-Krah
- & Colin W. Devey
-
Letter |
Dynamics of continental accretion
Three-dimensional dynamic computer models show how accretionary tectonic margins evolve from the initial plate-collision state, through a period of plate margin instability, and then re-establish a stable convergent margin; the models illustrate how significant curvature of the orogenic system develops, as well as the mechanism for tectonic escape of the back-arc region.
- L. Moresi
- , P. G. Betts
- & R. A. Cayley
-
Letter |
Lifespan of mountain ranges scaled by feedbacks between landsliding and erosion by rivers
Computational simulations show that variations in the rate of fluvial erosion between tectonically active and inactive mountain ranges may relate to a bidirectional coupling between bedrock river incision and landslides.
- David L. Egholm
- , Mads F. Knudsen
- & Mike Sandiford
-
Books & Arts |
Q&A: Species futurologist
Media artist Jon McCormack uses computer algorithms to imagine the future of native Australian species. As he prepares two new works — Codeform and Fifty Sisters — for the Ars Electronica Festival in Linz, Austria, he talks about digital evolution and virtual ecosystems.
- Jascha Hoffman
-
Books & Arts |
Q&A: The sound catcher
Tom Mitchell uses engineering and computing to enable people to play and sample live music using gestures. With the latest version of his co-creation 'The Gloves' about to debut at TEDGlobal 2012 in Edinburgh, UK, he talks about adaptive musical interaction.
- Jascha Hoffman
-
Letter |
Multiscale gigapixel photography
The AWARE-2 camera uses a parallel array of microcameras to capture one-gigapixel images at three frames per minute.
- D. J. Brady
- , M. E. Gehm
- & S. D. Feller
-
Books & Arts |
Q&A: The data visualizer
Aaron Koblin, head of the Data Arts Team in Google's Creative Lab, uses data visualization and crowdsourcing to reveal the changing relationship between people and technology. As he presents his work at the Eyeo Festival of digital creativity and prepares to release a collaboration with Google, London's Tate Modern and artist Chris Milk, he talks about the beauty of big data.
- Jascha Hoffman
-
Feature |
A lab app for that
Having a mobile device can save researchers a huge amount of time. It can also mean that they never switch off.
- Kendall Powell
-
Letter |
Engineered two-dimensional Ising interactions in a trapped-ion quantum simulator with hundreds of spins
A trapped-ion quantum simulator is used to demonstrate tunable long-range spin-spin couplings in two dimensions, relevant to studies of quantum magnetism at a scale that is intractable for classical computers.
- Joseph W. Britton
- , Brian C. Sawyer
- & John J. Bollinger
-
-
-
Books & Arts |
Q&A: The origami geometer
Computer scientist Erik Demaine uses origami to advance computational geometry and create art. His paper sculptures, made with his father, artist Martin Demaine, are now on show at the Japanese American National Museum in Los Angeles, California; from August, the exhibition will tour the United States. He explains the challenges of folding together mathematics and art.
- Jascha Hoffman
-
News |
Gamers outdo computers at matching up disease genes
Computer game crowdsources DNA sequence alignments across different species.
- Stephen Strauss