Fast Neural ODE / UDE: Improved Parallelism and Memory Performance Differentiating Stiff ODEs

Автор: Chris Rackauckas

Загружено: 2023-08-24

Просмотров: 1127

Описание: From ICIAM 2023 August 24th

https://iciam2023.org/registered_data...

[03338] Improved Parallelism and Memory Performance for Differentiating Stiff Differential Equations

Christopher Vincent Rackauckas (Julia Hub, Pumas-AI, MIT)

Abstract : Previous work demonstrated trade-offs in performance, numerical stability, and memory usage for ODE solving and differentiation of solutions. Our new time stepping methods expose more parallelism is shown to accelerate small ODE solves, while new GPU-based ODE solvers demonstrate a 10x performance improvement over Jax and PyTorch-based solvers. New adjoint methods achieve linear cost scaling with respect to parameters in stiff ODEs, as opposed to the cubic of Jax/PyTorch, while limiting the memory scaling.

I was on vacation so I had fun making that thumbnail. Isn't it awesome? I just Youtube'd the snizz out of that algorithm.

More information on the nonlinear solvers: https://sciml.ai/news/2024/01/23/nlso...

Не удается загрузить Youtube-плеер. Проверьте блокировку Youtube в вашей сети.
Повторяем попытку...

Fast Neural ODE / UDE: Improved Parallelism and Memory Performance Differentiating Stiff ODEs

Доступные форматы для скачивания:

Скачать видео

Информация по загрузке:

Скачать аудио

Похожие видео

Нейронные ОДУ (НОДУ) [Машинное обучение с учетом физики]

Нейронные ОДУ (НОДУ) [Машинное обучение с учетом физики]

David Emil Reich, Professor of Genetics, Harvard Medical School

David Emil Reich, Professor of Genetics, Harvard Medical School

Scientific Machine Learning and Stiffness - MIT Institute for AI and Fundamental Interactions IAIFI

Scientific Machine Learning and Stiffness - MIT Institute for AI and Fundamental Interactions IAIFI

Понимание Z-преобразования

Понимание Z-преобразования

Model Discovery w/ Imposed Structures and Prior Knowledge Scientific Machine Learning | ML4Science

Model Discovery w/ Imposed Structures and Prior Knowledge Scientific Machine Learning | ML4Science

Lec 2: Preferences and Utility Function

Lec 2: Preferences and Utility Function

Universal Differential Equations for Scientific Machine Learning - Chris Rackauckas MIT

Universal Differential Equations for Scientific Machine Learning - Chris Rackauckas MIT

ЛЕКЦИЯ ПРО НАДЁЖНЫЕ ШИФРЫ НА КОНФЕРЕНЦИИ БАЗОВЫХ ШКОЛ РАН В ТРОИЦКЕ

ЛЕКЦИЯ ПРО НАДЁЖНЫЕ ШИФРЫ НА КОНФЕРЕНЦИИ БАЗОВЫХ ШКОЛ РАН В ТРОИЦКЕ

Wyjaśniamy o co chodzi z Grenlandią. Czy naprawdę może wybuchnąć wojna USA-Dania?

Wyjaśniamy o co chodzi z Grenlandią. Czy naprawdę może wybuchnąć wojna USA-Dania?

Lecture 27: Current-Mode Control

Lecture 27: Current-Mode Control

What is machine learning and how can it be connected to prior scientific knowledge (SciML)?

What is machine learning and how can it be connected to prior scientific knowledge (SciML)?

Lecture 2: Analysis Methods and Rectifiers

Lecture 2: Analysis Methods and Rectifiers

12. Graphs, Networks, Incidence Matrices

12. Graphs, Networks, Incidence Matrices

The Numerical Analysis of Differentiable Simulation: Automatic Differentiation Can Be Incorrect

The Numerical Analysis of Differentiable Simulation: Automatic Differentiation Can Be Incorrect

What is (scientific) machine learning? An introduction through Julia's Lux.jl

What is (scientific) machine learning? An introduction through Julia's Lux.jl

Lecture 5: IS-LM Model

Lecture 5: IS-LM Model

Объяснение Transformers: понимание модели, лежащей в основе GPT, BERT и T5

Объяснение Transformers: понимание модели, лежащей в основе GPT, BERT и T5

MIT 6.S087: Foundation Models & Generative AI. CHAT-GPT & LLMs

MIT 6.S087: Foundation Models & Generative AI. CHAT-GPT & LLMs

Extending Scientific Machine Learning (SciML) to Agent-Based Models (ICLR AI4ABM 2023)

Extending Scientific Machine Learning (SciML) to Agent-Based Models (ICLR AI4ABM 2023)

Lec 12: Monopoly II

Lec 12: Monopoly II