東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Accelerating Atomistic Simulations W...

Johansson, Anders,

Linked to FindBook

Google Book

Amazon

博客來

Accelerating Atomistic Simulations With GPUs and Machine Learning /

Record Type:	Electronic resources : Monograph/item
Title/Author:	Accelerating Atomistic Simulations With GPUs and Machine Learning // Anders Johansson.
Author:	Johansson, Anders,
Description:	1 electronic resource (164 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 85-12, Section: B.
Contained By:	Dissertations Abstracts International85-12B.
Subject:	Computational physics. -
Online resource:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31296998
ISBN:	9798382784632

Accelerating Atomistic Simulations With GPUs and Machine Learning /
Johansson, Anders,

Accelerating Atomistic Simulations With GPUs and Machine Learning /Anders Johansson. - 1 electronic resource (164 pages)

Source: Dissertations Abstracts International, Volume: 85-12, Section: B.

The field of atomistic simulations is rapidly evolving. The simulations are becoming more reliable and accurate through the invention of ever more advanced ML potentials such as uncertainty aware and equivariant methods. Simultaneously, the leading GPU-based supercomputers are finally reaching the exascale, which paves the way for larger and longer simulations. This thesis revolves around the intersection of these advances and the acceleration of state-of-the-art ML potentials with modern GPUs on the world's largest supercomputers.More concretely, my focus has been the performance portability and scalability of FLARE and Allegro, two ML potentials near the opposite ends of the cost-accuracy trade-off. FLARE is a sparse Gaussian process potential that attempts to push the envelope for speed while maintaining reasonable accuracy. I have developed a Kokkos implementation of FLARE that outperformed previous state-of-the-art methods by 70% on the second fastest supercomputer in the world, and recently reached one trillion atoms on Frontier, the fastest supercomputer. Allegro is an equivariant neural network implemented in PyTorch which sacrifices some speed to reach leading accuracy while maintaining scalability through its innovative architecture. Through the scalability and my LAMMPS interface, Allegro was able to efficiently utilize 5120 GPUs and reach relevant speeds for a wide range of biomoleular structures.

English

ISBN: 9798382784632Subjects--Topical Terms:

3343998
Computational physics.
Subjects--Index Terms:

Active learning

Accelerating Atomistic Simulations With GPUs and Machine Learning /
LDR:02947nmm a22004693i 4500 001 2433736
005 20250905091002.5
006 m o d
007 cr|nu||||||||
008 261011s2024 miu||||||m |||||||eng d
020 $a 9798382784632
035 $a (MiAaPQD)AAI31296998
035 $a AAI31296998
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Johansson, Anders, $e author. $0 (orcid)0000-0002-0544-7202 $3 3823474
245 1 0 $a Accelerating Atomistic Simulations With GPUs and Machine Learning / $c Anders Johansson.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2024
300 $a 1 electronic resource (164 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 85-12, Section: B.
500 $a Advisors: Kozinsky, Boris Committee members: Brenner, Michael; Clarke, David; Spaepen, Frans.
502 $b Ph.D. $c Harvard University $d 2024.
520 $a The field of atomistic simulations is rapidly evolving. The simulations are becoming more reliable and accurate through the invention of ever more advanced ML potentials such as uncertainty aware and equivariant methods. Simultaneously, the leading GPU-based supercomputers are finally reaching the exascale, which paves the way for larger and longer simulations. This thesis revolves around the intersection of these advances and the acceleration of state-of-the-art ML potentials with modern GPUs on the world's largest supercomputers.More concretely, my focus has been the performance portability and scalability of FLARE and Allegro, two ML potentials near the opposite ends of the cost-accuracy trade-off. FLARE is a sparse Gaussian process potential that attempts to push the envelope for speed while maintaining reasonable accuracy. I have developed a Kokkos implementation of FLARE that outperformed previous state-of-the-art methods by 70% on the second fastest supercomputer in the world, and recently reached one trillion atoms on Frontier, the fastest supercomputer. Allegro is an equivariant neural network implemented in PyTorch which sacrifices some speed to reach leading accuracy while maintaining scalability through its innovative architecture. Through the scalability and my LAMMPS interface, Allegro was able to efficiently utilize 5120 GPUs and reach relevant speeds for a wide range of biomoleular structures.
546 $a English
590 $a School code: 0084
650 4 $a Computational physics. $3 3343998
650 4 $a Computer science. $3 523869
650 4 $a Materials science. $3 543314
650 4 $a Engineering. $3 586835
653 $a Active learning
653 $a Equivariance
653 $a Atomistic simulations
653 $a Machine learning
653 $a Molecular dynamics
653 $a Performance portability
690 $a 0216
690 $a 0794
690 $a 0984
690 $a 0800
690 $a 0537
710 2 $a Harvard University. $b Engineering and Applied Sciences - Applied Physics. $e degree granting institution. $3 3823459
720 1 $a Kozinsky, Boris $e degree supervisor.
773 0 $t Dissertations Abstracts International $g 85-12B.
790 $a 0084
791 $a Ph.D.
792 $a 2024
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31296998