JuliaCon 2020 | Doing Scientific Machine Learning (SciML) With Julia
Scientific machine learning combines differentiable programming, scientific simulation (differential equations, nonlinear solvers, etc.), and machine learning (deep learning) in order impose physical constraints on machine learning and automatically learn biological models. Given the composibility of Julia, many have noted that it is positioned as the best language for this set of numerical techniques, but how to do actually "do" SciML? This workshop gets your hands dirty.
In this workshop we'll dive into some of the latest techniques in scientific machine learning, including Universal Differential Equations (Universal Differential Equations for Scientific Machine Learning), Physics-Informed Neural Networks (Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations), and Sparse Identification of Nonlinear Dynamics (SInDy, Discovering governing equations from data by sparse identification of nonlinear dynamical systems). The goal is to get those in the workshop familiar with what these methods are, what kinds of problems they solve, and know how to use Julia packages to implement them.
The workshop will jump right into how to model the missing part of a physical simulation, describe how universal approximators (neural networks) can be used in this context, and show how to transform such problems into an optimization problem which is then accelerated by specializing automatic differentiation. The set of packages that is involved in this is somewhat intense, using many tools from JuliaDiffEq (DiffEqFlux.jl, DifferentialEquations.jl, DiffEqSensitivity.jl, ModelingToolkit.jl, NeuralPDE.jl, DataDrivenDiffEq.jl, Surrogates.jl, etc.) combined with machine learning tools (Flux.jl), differentiation tooling (SparseDiffTools.jl, Zygote.jl, ForwardDiff.jl, ReverseDiff.jl, etc.), and optimization tooling (JuMP, Optim.jl, Flux.jl, NLopt.jl, etc.) all spun together in a glorious soup that automatically discovers physical laws at the end of the day. Thus this workshop has something different to offer for everyone: new users of Julia will get a nice overview of the unique composibility of the Julia package ecosystem, while experienced Julia users will learn how to bridge some area that they are comfortable with (such as machine learning) to a whole new set of phenomena. Meanwhile, even though who are only knee deep in coding can gain a lot from learning these new mathematical advances, meaning that even a casual observer likely has a lot to learn!
Slides: https://figshare.com/articles/presentation/Doing_Scientific_Machine_Learning_with_Julia_s_SciML_Ecosystem/12751949
Torchdiffeq benchmarks: https://gist.github.com/ChrisRackauckas/cc6ac746e2dfd285c28e0584a2bfd320
Neural ODE training benchmarks: https://gist.github.com/ChrisRackauckas/4a4d526c15cc4170ce37da837bfc32c4
torchsde benchmarks: https://gist.github.com/ChrisRackauckas/6a03e7b151c86b32d74b41af54d495c6
Видео JuliaCon 2020 | Doing Scientific Machine Learning (SciML) With Julia канала The Julia Programming Language
In this workshop we'll dive into some of the latest techniques in scientific machine learning, including Universal Differential Equations (Universal Differential Equations for Scientific Machine Learning), Physics-Informed Neural Networks (Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations), and Sparse Identification of Nonlinear Dynamics (SInDy, Discovering governing equations from data by sparse identification of nonlinear dynamical systems). The goal is to get those in the workshop familiar with what these methods are, what kinds of problems they solve, and know how to use Julia packages to implement them.
The workshop will jump right into how to model the missing part of a physical simulation, describe how universal approximators (neural networks) can be used in this context, and show how to transform such problems into an optimization problem which is then accelerated by specializing automatic differentiation. The set of packages that is involved in this is somewhat intense, using many tools from JuliaDiffEq (DiffEqFlux.jl, DifferentialEquations.jl, DiffEqSensitivity.jl, ModelingToolkit.jl, NeuralPDE.jl, DataDrivenDiffEq.jl, Surrogates.jl, etc.) combined with machine learning tools (Flux.jl), differentiation tooling (SparseDiffTools.jl, Zygote.jl, ForwardDiff.jl, ReverseDiff.jl, etc.), and optimization tooling (JuMP, Optim.jl, Flux.jl, NLopt.jl, etc.) all spun together in a glorious soup that automatically discovers physical laws at the end of the day. Thus this workshop has something different to offer for everyone: new users of Julia will get a nice overview of the unique composibility of the Julia package ecosystem, while experienced Julia users will learn how to bridge some area that they are comfortable with (such as machine learning) to a whole new set of phenomena. Meanwhile, even though who are only knee deep in coding can gain a lot from learning these new mathematical advances, meaning that even a casual observer likely has a lot to learn!
Slides: https://figshare.com/articles/presentation/Doing_Scientific_Machine_Learning_with_Julia_s_SciML_Ecosystem/12751949
Torchdiffeq benchmarks: https://gist.github.com/ChrisRackauckas/cc6ac746e2dfd285c28e0584a2bfd320
Neural ODE training benchmarks: https://gist.github.com/ChrisRackauckas/4a4d526c15cc4170ce37da837bfc32c4
torchsde benchmarks: https://gist.github.com/ChrisRackauckas/6a03e7b151c86b32d74b41af54d495c6
Видео JuliaCon 2020 | Doing Scientific Machine Learning (SciML) With Julia канала The Julia Programming Language
Показать
Комментарии отсутствуют
Информация о видео
26 июля 2020 г. 23:10:01
03:58:13
Другие видео канала
![JuliaCon 2020 | A Deep Dive into DataFrames.jl Indexing](https://i.ytimg.com/vi/X7AJf0C4Czw/default.jpg)
![BuzzConf 2020 - Chris Rackauckas - SciML: How Language is Changing Scientific Research](https://i.ytimg.com/vi/3hM7wn7iJ70/default.jpg)
![Convolutions in image processing | Week 1 | MIT 18.S191 Fall 2020 | Grant Sanderson](https://i.ytimg.com/vi/8rrHTtUzyZA/default.jpg)
![Paper towns and why learning is awesome | John Green](https://i.ytimg.com/vi/NgDGlcxYrhQ/default.jpg)
![JuliaCon 2020 | Keynote: Scientific Machine Learning | Prof Karen Willcox](https://i.ytimg.com/vi/Bk4PJnjuPps/default.jpg)
![JuliaCon 2020 | Using VS Code for Julia development | David Anthoff](https://i.ytimg.com/vi/IdhnP00Y1Ks/default.jpg)
![Scientific Machine Learning and Stiffness - MIT Institute for AI and Fundamental Interactions](https://i.ytimg.com/vi/FENK1SDvPiA/default.jpg)
![ICTP-EAIFR Colloquium on "Machine learning and molecular dynamics"](https://i.ytimg.com/vi/cAhn4Z3Rv9M/default.jpg)
![AI4Pandemics Talk #1: Learning Epidemic Models That Extrapolate, Chris Rackauckas. July 14, 2021.](https://i.ytimg.com/vi/7yPU_04Unb8/default.jpg)
![Rust, Julia, and Go: Disruptive New Programming Languages Changing the Face of Computing](https://i.ytimg.com/vi/6qCH7Y2rc_w/default.jpg)
![JuliaCon 2020 | Manifolds in Julia – Manifolds.jl & ManifoldsBase.jl | Ronny Bergmann](https://i.ytimg.com/vi/md-FnDGCh9M/default.jpg)
![What is Automatic Differentiation?](https://i.ytimg.com/vi/wG_nF1awSSY/default.jpg)
![Parallel Computing on Your Own Machine | Week 8 | 18.S191 MIT Fall 2020](https://i.ytimg.com/vi/dczkYlOM2sg/default.jpg)
![10 Julia Packages You Should Learn for Data Science (in 2020)](https://i.ytimg.com/vi/ciD0IqQ5yIA/default.jpg)
![Scientific Machine Learning: Where Physics-based Modeling Meets Data-driven Learning](https://i.ytimg.com/vi/JgwwDvvOHk4/default.jpg)
![Julia Ling: "Machine Learning for Materials Discovery" | IACS Seminar](https://i.ytimg.com/vi/28Ue_jteKI4/default.jpg)
![13 Machine Learning: Time Series Analysis](https://i.ytimg.com/vi/nxY1HS3SrQ4/default.jpg)
![Chris Rackauckas - Physics-Informed Neural Networks (PINNs) | Podcast #42](https://i.ytimg.com/vi/OmySUTFwh2g/default.jpg)
![Beyond Deep Learning - Differentiable Programming with Flux - Avik Sengupta | ODSC Europe 2019](https://i.ytimg.com/vi/FZ1wlzFp0r8/default.jpg)
![JuliaCon 2020 | State of Julia | Jeff Bezanson & Stefan Karpinski](https://i.ytimg.com/vi/xKrIp4ZVOrg/default.jpg)