Papers of the day   All papers

Newton vs the machine: solving the chaotic three-body problem using deep neural networks

Comments

Oct 26 2019 Eric Topol

In 1687 Sir Issac Newton posited the three-body problem in Principia. 332 years later it was solved with a deep neural network, and 100 million times faster https://www.technologyreview.com/s/614597/a-neural-net-solves-the-three-body-problem-100-million-times-faster/ @techreview #AI https://arxiv.org/pdf/1910.07291.pdf Philip Breen @EdinburghUni cc @stevenstrogatz for input https://t.co/vYC3plj9Jp
5 replies, 223 likes


Oct 26 2019 Kovas Boguta

"Complex systems winter" is coming to a close. Newton vs the machine: solving the chaotic three-body problem using deep neural networks https://arxiv.org/abs/1910.07291
1 replies, 18 likes


Oct 26 2019 Daisuke Okanohara

The three-body problem has no analytic solution in general, and its simulation cost tends to be very large due to its chaotic nature. Surprisingly, NN can approximate the solution at a fixed cost and up to 100 million times faster than existing solvers. https://arxiv.org/abs/1910.07291
0 replies, 12 likes


Oct 17 2019 Joshua Bloom

Funny figure aside, I enjoyed this paper out today: "Newton vs the machine: solving the chaotic three-body problem using deep neural networks" by Breen et al. Great example of cross-disciplinary work between N-body astronomers and computational folks https://arxiv.org/abs/1910.07291 https://t.co/Q5JyEsOnxl
1 replies, 9 likes


Oct 28 2019 Carlos E. Perez 🧢

Deep Learning solves the 3-body problem! If you aren't convinced yet, then you are beyond redemption!! https://arxiv.org/abs/1910.07291 #ai #deeplearning
1 replies, 8 likes


Oct 26 2019 Umut Eser

Super sensitivity to micro states is often prohibitive for bottom up knowledge-driven simulations. But thanks to emergent behaviours, there are also a huge ton of invariences which can be modeled with data driven inferences (e.g. Deep Learning). 1/2
1 replies, 5 likes


Oct 29 2019 Sibesh Kar

Misleading. Any solution that involves numerical methods always requires the same amount of compute, regardless of whether that compute is used to calculate approximate trajectories, or train neural networks to approximate trajectories. Can't escape limits of information theory
0 replies, 4 likes


Oct 17 2019 philipbreen

Check out my latest article: Newton vs the machine: using deep neural networks to tackle a 300-year-old problem https://www.linkedin.com/pulse/newton-vs-machine-using-deep-neural-networks-tackle-problem-breen via @LinkedIn or the preprint on https://arxiv.org/abs/1910.07291 @arxiv
0 replies, 3 likes


Oct 27 2019 Pascal Kwanten | Less Artificial Very Intelligist

https://arxiv.org/abs/1910.07291
0 replies, 2 likes


Oct 19 2019 Machine Learning: Science and Technology

Newton vs the machine: solving the chaotic three-body problem using deep neural networks at https://arxiv.org/abs/1910.07291 #MachineLearning #astrophysics
0 replies, 2 likes


Oct 27 2019 Carlos Ciller

Early morning #MachineLearning digest - Newton vs The Machine: Solving The Chaotic Three-Body Problem Using Deep Neural Networks via @techreview #breakfast Article https://www.technologyreview.com/s/614597/a-neural-net-solves-the-three-body-problem-100-million-times-faster/ #Arxiv Paper http://arxiv.org/abs/1910.07291
0 replies, 2 likes


Oct 26 2019 Larry Hunter

Cool result: "A deep neural network ... provides accurate solutions at fixed computational cost up to 100 million times faster than state-of-the-art ... numerical solvers, enabling fast and scalable simulations of many-body systems.” https://arxiv.org/pdf/1910.07291.pdf
0 replies, 2 likes


Oct 17 2019 Machine Learning

Newton vs the machine: solving the chaotic three-body problem using deep neural networks. http://arxiv.org/abs/1910.07291
0 replies, 2 likes


Nov 01 2019 DLビギナ

https://twitter.com/AkiraTOSEI/status/1189882225451397120?16814
0 replies, 1 likes


Content