The Julia SciML Ecosystem: Scientific Machine Learning as a Software Problem - Chris Rackauckas
The Julia SciML Ecosystem: Scientific Machine Learning as a Software Problem
Christopher V. Rackauckas, Massachusetts Institute of Technology
Abstract:
The Julia SciML ecosystem is a software suite for high performance modeling and simulation which incorporates scientific machine learning for automated model discovery and nonlinear model order reduction. This talk will introduce the audience to the philosophy of composability of the software suite by demonstrating how connections between seemingly disjoint modules can be composed to create new high performance algorithms on demand. It will start by describing the high performance simulation techniques, from new methods for numerically approximating ODEs and SDEs to all the way to methods for generating discretizations for nonlinear optimal control and automated solution of partial differential equations. From there connections to machine learning will be seen, with physics-informed neural networks (PINNs) supplementing the suite with non-local partial differential-algebraic equation solvers to universal differential equations with the ability to learn model misspecification directly from data. Connections to probabilistic programming will be discussed, mixing Bayesian estimation with scientific machine learning, along with high performance computing via CPU and GPU parallelism. This talk will thus set the stage for the following speakers who will demonstrate specific methodologies and applications within this application sphere.
For more information on the SciML Open Source Scientific Machine Learning Software Ecosystem, see https://sciml.ai/.
For more info on the Julia Programming Language, follow us on Twitter: https://twitter.com/JuliaLanguage
Contents
00:00 Overview
00:21 Objective: The aim of the SciML ecosystem
01:24 Definition: What is scientific machine learning
02:45 Example: An application to ocean columns modelling
07:25 Hard Problem: How to fit a neural network inside a simulator
10:59 No Silver Bullet: Different adjoint methods for different problems
13:24 SciML vs. The Rest: A comparison with other libraries in other languages
14:38 SciML Today: The current state of the SciML ecosystem
15:35 Conclusion: Further developments
S/O to https://github.com/pitmonticone for the video timestamps!
Want to help add timestamps to our YouTube videos to help with discoverability? Find out more here: https://github.com/JuliaCommunity/YouTubeVideoTimestamps
Interested in improving the auto generated captions? Get involved here: https://github.com/JuliaCommunity/YouTubeVideoSubtitles
Видео The Julia SciML Ecosystem: Scientific Machine Learning as a Software Problem - Chris Rackauckas канала The Julia Programming Language
Christopher V. Rackauckas, Massachusetts Institute of Technology
Abstract:
The Julia SciML ecosystem is a software suite for high performance modeling and simulation which incorporates scientific machine learning for automated model discovery and nonlinear model order reduction. This talk will introduce the audience to the philosophy of composability of the software suite by demonstrating how connections between seemingly disjoint modules can be composed to create new high performance algorithms on demand. It will start by describing the high performance simulation techniques, from new methods for numerically approximating ODEs and SDEs to all the way to methods for generating discretizations for nonlinear optimal control and automated solution of partial differential equations. From there connections to machine learning will be seen, with physics-informed neural networks (PINNs) supplementing the suite with non-local partial differential-algebraic equation solvers to universal differential equations with the ability to learn model misspecification directly from data. Connections to probabilistic programming will be discussed, mixing Bayesian estimation with scientific machine learning, along with high performance computing via CPU and GPU parallelism. This talk will thus set the stage for the following speakers who will demonstrate specific methodologies and applications within this application sphere.
For more information on the SciML Open Source Scientific Machine Learning Software Ecosystem, see https://sciml.ai/.
For more info on the Julia Programming Language, follow us on Twitter: https://twitter.com/JuliaLanguage
Contents
00:00 Overview
00:21 Objective: The aim of the SciML ecosystem
01:24 Definition: What is scientific machine learning
02:45 Example: An application to ocean columns modelling
07:25 Hard Problem: How to fit a neural network inside a simulator
10:59 No Silver Bullet: Different adjoint methods for different problems
13:24 SciML vs. The Rest: A comparison with other libraries in other languages
14:38 SciML Today: The current state of the SciML ecosystem
15:35 Conclusion: Further developments
S/O to https://github.com/pitmonticone for the video timestamps!
Want to help add timestamps to our YouTube videos to help with discoverability? Find out more here: https://github.com/JuliaCommunity/YouTubeVideoTimestamps
Interested in improving the auto generated captions? Get involved here: https://github.com/JuliaCommunity/YouTubeVideoSubtitles
Видео The Julia SciML Ecosystem: Scientific Machine Learning as a Software Problem - Chris Rackauckas канала The Julia Programming Language
Показать
Комментарии отсутствуют
Информация о видео
9 марта 2021 г. 21:11:01
00:16:10
Другие видео канала
![EuroAD 2021: ChainRules.jl: AD system agnostic rules for JuliaLang](https://i.ytimg.com/vi/B3bC49OmTdk/default.jpg)
![Keynote: Modeling and Simulation to Guide Dose... | Husain Attarwala (Moderna) | JuliaCon 2022](https://i.ytimg.com/vi/6wGSCD3cI9E/default.jpg)
![UnitCommitment.jl Security-Constrained Unit Commitment in JuMP | Alinson S Xavier | JuliaCon2021](https://i.ytimg.com/vi/rYUZK9kYeIY/default.jpg)
![JuliaCon 2020 | Convex.jl: where are we and where do we want to go? | Eric P. Hanson](https://i.ytimg.com/vi/iczcGK6G_sE/default.jpg)
![Put some constraints into your life with JuliaCon(straints) | Jean-François Baffier | JuliaCon 2021](https://i.ytimg.com/vi/G4siuvNMj0c/default.jpg)
![The OSCAR Computer Algebra System | Max Horn, Claus Fieker | JuliaCon 2021](https://i.ytimg.com/vi/rWQK4mU3jQc/default.jpg)
![3.6x speedup on A64FX by squeezing ShallowWaters.jl into Float16 | Milan Klöwer | JuliaCon2021](https://i.ytimg.com/vi/btHfZr2C0GA/default.jpg)
![Relational AI | Sponsor Talk | JuliaCon 2021](https://i.ytimg.com/vi/BTmmjTAwNgI/default.jpg)
![Pebble games - Time and space to differentiate a program | Jin Guo Lin | JuliaCon2021](https://i.ytimg.com/vi/A9uMxPST_HE/default.jpg)
![AugmentedGaussianProcesses.jl, a full Gaussian Process toolkit | Théo Galy-Fajou | JuliaCon 2020](https://i.ytimg.com/vi/MLDijVso6Vk/default.jpg)
![Introduction to Decision Making Under Uncertainty using POMDPs.jl](https://i.ytimg.com/vi/yrC2O7T7aKU/default.jpg)
![Julog.jl: Prolog-like Logic Programming in Julia | Xuan (Tan Zhi Xuan) | JuliaCon 2021](https://i.ytimg.com/vi/FzHML1bbcSU/default.jpg)
![Bifurcation Based Machine Learning of Dynamical Systems | Kyoung Hyun Lee | SciMLCon 2022](https://i.ytimg.com/vi/O-BbP7W6kvk/default.jpg)
![RelationalAI | Sponsored Talk | JuliaCon 2022](https://i.ytimg.com/vi/Pncp0puIwI0/default.jpg)
![Agents.jl and the next chapter in agent based modelling | Tim DuBois | JuliaCon 2021](https://i.ytimg.com/vi/Iaco6v6TVXk/default.jpg)
![Tomographic Image Reconstruction with Julia | Tobias Knopp | JuliaCon2021](https://i.ytimg.com/vi/cDZq_vr10CQ/default.jpg)
![Learn about Blockchain Development in Julia | Logan Kilpatrick | JuliaCon2021](https://i.ytimg.com/vi/1OgxuV1LsQo/default.jpg)
![Scaling up Training of any Flux.jl Model Made Easy | Dhairya Gandhi | JuliaCon 2022](https://i.ytimg.com/vi/176c4S6LqT8/default.jpg)
![Sampling Live Visualizations with Turkie and TuringCallbacks | Théo Galy-Fajou | JuliaCon2021](https://i.ytimg.com/vi/Tu9GRus7-FM/default.jpg)
![Jumping into the Julia Community via Advent Of Code | Jasmine Hughes | JuliaCon2021](https://i.ytimg.com/vi/336gAu7PAlI/default.jpg)
![Systems Biology in ModelingToolkit | A Jain, S Iravanian, P Lang | JuliaCon2021](https://i.ytimg.com/vi/DL0Xw7ETZsE/default.jpg)