Papers of the day   All papers

Fourier Neural Operator for Parametric Partial Differential Equations


Prof. Anima Anandkumar: Fourier neural operator for PDEs Solves family of #PDE from scratch at any resolution. Outperforms all existing #DeepLearning methods. 1000x faster than traditional solvers Experiments on Navier-Stokes equations @kazizzad @Caltech #AI #HPC

30 replies, 1908 likes

Prof. Anima Anandkumar: This is wild! @MCHammer retweeted our work on Fourier neural operator with @ZongyiLiCaltech @kazizzad it can solve Navier Stokes and other PDEs from scratch. Thanks @MCHammer for using your influence for popularizing science #scicomm #AI #DeepLearning

7 replies, 429 likes

MIT CSAIL: Fourier neural operator for PDEs ⚬ solves family of #PDE from scratch at any resolution ⚬ outperforms existing deep-learning methods Paper: (v/@Caltech) #AI #ML #DeepLearning

1 replies, 374 likes

math prof: Interesting ... I take that back REAL interesting.

1 replies, 327 likes

Prof. Anima Anandkumar: If you are excited about #AI research we are doing you need to support more #WomenInSTEM #immigrants #BLM #VoteBidenHarrisToSaveAmerica #Vote2020

5 replies, 214 likes

Prof. Anima Anandkumar: Blog post by @ZongyiLiCaltech Nice summary of work and results. @kazizzad @Caltech

0 replies, 200 likes

Prof. Anima Anandkumar: Thanks for retweeting our work @MIT_CSAIL @kazizzad @ZongyiLiCaltech

1 replies, 183 likes

Prof. Anima Anandkumar: [email protected] can you unlock @ZongyiLiCaltech After @MCHammer retweeted our work @ZongyiLiCaltech newly created account got a lot of following and retweet. Your dumb method wrongly classified it as spam. You need to give verified status to researchers. I still don't have one

6 replies, 156 likes

Peter Wang: MC Hammer is right. This application of deep neutral nets to PDEs is really cool!

5 replies, 80 likes

Ryan Abernathey: Another unexpected development from 2020: learning about cool new ML methods for physics simulation from MC Hammer. 🤘

2 replies, 66 likes

Robert Fisher: Amazing new progress on solving key PDEs, including the Navier-Stokes equation, using #DeepLearning. At first it may seem surprising that AI can tackle these time-dependent complex flows. At high Reynolds numbers, however, Navier-stokes is hyperbolic. (1/3)

4 replies, 64 likes

Kamyar Azizzadenesheli: @shekar09 @AnimaAnandkumar @Caltech Hi there, These series of papers: among many in their bibs, and related ones may be useful. We tried to make the first paper very accessible ;).

0 replies, 62 likes

Orazio Gallo: My new career goal is to write a paper that gets tweeted by @MCHammer, like @AnimaAnandkumar and her collaborators did.

0 replies, 52 likes

ML and Data Projects To Know: 📙 Title: Fourier Neural Operator for Parametric Partial Differential Equations Authors: Zongyi Li, Nikola Kovachki, @kazizzad Burigede Liu, Kaushik Bhattacharya, Andrew Stuart, @AnimaAnandkumar Featured in PTK #58 Paper:

0 replies, 47 likes

ESS: It's 2020, you wake up to check your Twitter feed and MC Hammer is posting bleeding edge machine learning papers. You're ok with this timeline.

0 replies, 44 likes

Dan Kaminsky: This will solve new things.

4 replies, 40 likes

Popular ML resources: The most popular ArXiv tweet in the last 24h:

0 replies, 30 likes

Kamyar Azizzadenesheli: Finally, this paper is out 🙋‍♂️, learning operators is fun, and Fourier is back to help us to enjoy it more.

1 replies, 26 likes

Sam Kimbinyi: MC Hammer posting Machine Learning papers is exactly what 2020 needed

1 replies, 24 likes

HotComputerScience: Most popular computer science paper of the day: "Fourier Neural Operator for Parametric Partial Differential Equations"

0 replies, 21 likes

Zachary Ross: This is amazing

0 replies, 20 likes

Bjorn Heijligers: Deep neural nets have been used to speed up N-body calculations in cosmological simulations. Now also for PDE's of fluid dynamics. 1000x speed increase. Arthur C. Clarke's 3rd law applies. Any sufficiently advanced technology is indistinguisable from magic.

0 replies, 19 likes

Grady Booch: U Can TensorFlow That

0 replies, 19 likes

Deep Prasad: Amazing research on solving families of partial differential equations using neural operator based neural networks. One level of abstraction beyond solving individual PDEs themselves. Cool!!Will be interesting to see how this helps CFD simulations.Thanks for posting it MC Hammer.

2 replies, 18 likes

Dave Lauer: Please Hammer, don't hurt em'

0 replies, 16 likes

Frank Odom: This just made my day. Thank you @MCHammer for spreading the gospel. #Science #DeepLearning #HammerTime

0 replies, 16 likes

Juanjo Garcia Ripoll: This is very cool. I am surprised just four layers give this accuracy for nonlinear equations, irrespective of the discretization. Sounds promising also for our efforts on MPS-based renormalization techniques.

0 replies, 15 likes

Thorsten Becker: Wow

2 replies, 15 likes

Cami Rosso: Interesting. A novel #neuralnetwork operator (for #PDE, can be used for #computervision) that parameterizes the integral kernel directly in #Fourier space—Applies activation functions on the spatial domain #innovation #Leadership #AI #MachineLearning #DeepLearning #NavierStokes

0 replies, 14 likes

Francisco Javier Arceo🇲🇽🇺🇸: This is awesome. I remember how amazed I was the first time I learned about #FourierTransforms. Seeming them applied to #DeepLearning and Partial Differential Equations is really exciting!!!

1 replies, 14 likes

Perspicuous: MC Hammer tweeting about PDEs. This is way more interesting than most anything else on twitter.

0 replies, 13 likes

dragosr: AI Pattern Matching in Fourier space rapidly predicts solutions for differential equations. Via, MC Hammer, yes really, and huge.

0 replies, 12 likes

Anand Jebakumar: Very interesting and promising work from @AnimaAnandkumar's lab. This might help move beyond traditional discretization schemes for CFD.

0 replies, 12 likes

brandon 🏳️‍🌈🗿: ??

2 replies, 11 likes

grounded maverick: A remarkable achievement in the field of fluid dynamics studies which will definitely change the course of the way fluid and turbulance simulations are done.

0 replies, 10 likes

Prof. Anima Anandkumar: And original tweet

0 replies, 10 likes

Disruption Joe: This is so much better than I ever expected 2020 to offer.

0 replies, 8 likes

Alan Myron : Science, Politics, SciFi News 💜🇺🇸💪🏼🌺: This is big.

0 replies, 7 likes

Jerome A Johnson: Hammer Time with solutions for partial differential equations. 😁

1 replies, 5 likes

LimpDickScienceMajor: Stop! It's solving PDE's via new Fourier neural operator time!

0 replies, 4 likes

Susheel Varma: Up to three orders of magnitude speedup of Navier-Stokes equations in the Fourier domain using Deep Learning!! Now I have seen everything!! Arthur C. Clarke's third law - Any sufficiently advanced technology is indistinguisable from magic.

0 replies, 3 likes

Daisuke Okanohara: As an extension of Neural Operator, Fourier Neural Operator parameterizes the integral kernel in Fourier space (when the kernel is shift-invariant). Achieving SOTAs accuracy and 1000x times faster than traditional PDE solvers.

0 replies, 3 likes

Michael Mindrum, MD: Maybe this can help your work in metabolism @KevinH_PhD? H/t to MC Hammer of course.

2 replies, 3 likes

Elco Luijendijk: This could be kind of a game changer for modelling fluid flow or any kind of geological process really. Three orders of magnitude speedup of models with neural networks compared to regular finite element models

0 replies, 3 likes

Michael Chapiro: People on clubhouse: "why should I care about rapper's or actor's takes on various things?" A rapper on twitter: *tweets about breakthrough in deep learning for solving PDEs before you* h/t @StephenFleming

0 replies, 3 likes

Miguel DJ: @MCHammer is one of my top science tweeters

0 replies, 3 likes

Rahul M Dodhia: This is actually a great paper: read the layman explanation:

0 replies, 3 likes

Oliver Lord: This could be really useful for geodynamic numerical calculations of many kinds: for the atmosphere and the Deep Earth. Despite the hype, #DeepLearning is an exciting approach to a many problems. Also, you know it’s a significant breakthrough when it’s retweeted by [email protected]

1 replies, 3 likes

Daniel Probst: Noice

0 replies, 2 likes

Jonah Miller: Interested in your thoughts on this @ArvindMohan15

0 replies, 2 likes

Ray Radlein: Let's check in on what MC Hammer is tweeting about:

0 replies, 2 likes

∂≈∞ --- ∫∫∫∫∫∫∫∫∫∫∫∫∫∫∫∫∆∏∑∫∫∫∫∫∫∫∫∫∫∫∫∫∫∫∫: #NeuralNetworks #ai #DeepLearning a great groundbreaking publication! Best thing I saw today all day on twitter.

0 replies, 2 likes 🔨⏰

0 replies, 1 likes

Mathias Niepert: And this is a PDE, uh, you can't solve You can’t touch it You can’t touch it My, my, my my deep network hits it so hard

0 replies, 1 likes

Texture: The weirdest timeline

0 replies, 1 likes

Thomas Analytics Int'l, LLC: #DataScientists #BigData #AI #RPA #IoT #Linux

0 replies, 1 likes

Hapless Dark Star: Very interesting approach to using Neural Nets to solve Partial Differential Equations by transforming into frequency domain.

1 replies, 1 likes

Dave Sag: Wow. MC Hammer is a machine learning nerd. I never knew that.

1 replies, 1 likes

Kenton Murray: Yes. This is actually MC Hammer tweeting about a paper on NNs and PDEs.

0 replies, 1 likes

off the shelf edge: Wow, this is a pretty big deal! Darcy flow in the subsurface is a great example of a Navier-Stokes equation that can benefit from this!

1 replies, 1 likes

Stephen Borstelmann MD: @Miles_Brundage @BrundageBot you've got competition.

0 replies, 1 likes

Billy Bob Bain:

0 replies, 1 likes

Matt Pettis: Getting AI content from @MCHammer is definitely the best "not on my 2020 bingo card" event this year. Plus, this

0 replies, 1 likes

Risav Karna: 2020 has hope. Also this is everything I needed neural networks to do...

0 replies, 1 likes


Found on Oct 21 2020 at

PDF content of a computer science paper: Fourier Neural Operator for Parametric Partial Differential Equations