|
|
Hokkaido University Collection of Scholarly and Academic Papers >
Information Initiative Center >
Peer-reviewed Journal Articles, etc >
A Cholesky QR type algorithm for computing tall-skinny QR factorization with column pivoting
| Title: | A Cholesky QR type algorithm for computing tall-skinny QR factorization with column pivoting |
| Authors: | Fukaya, Takeshi Browse this author →KAKEN DB | | Nakatsukasa, Yuji Browse this author | | Yamamoto, Yusaku Browse this author →KAKEN DB |
| Keywords: | Performance evaluation | | Distributed processing | | Accuracy | | High performance computing | | Approximation algorithms | | Iterative algorithms | | Computational efficiency |
| Issue Date: | 8-Jul-2024 |
| Publisher: | IEEE |
| Journal Title: | 2024 IEEE International Parallel and Distributed Processing Symposium (IPDPS) |
| Start Page: | 63 |
| End Page: | 75 |
| Publisher DOI: | 10.1109/IPDPS57955.2024.00015 |
| Abstract: | Innovative methods, new instruments, disruptive techniques, and groundbreaking technologies have led tosignificant leaps in scientific progress. The increasingly powerful Large Language Models (LLMs) releasedeach month already speed up research activities such as concept explanation, literature search, andsummarization. The transformative potential of AI in research activities, in particular foundation models,raises important questions about their performance in science activities, their potential application indifferent contexts, and their ethics. In this talk, I will introduce AuroraGPT, Argonne National Laboratory'seffort to explore the notion of AI research assistants. To illustrate the gap between existing LLMs and anideal AI research assistant, I will first share observations from using existing LLMs as early researchassistants in three parallel and distributed computing experiments with experts in scheduling, distributedprotocols, and PDE solvers. AuroraGPT is developed as an open foundation model trained specifically withscientific data to explore solutions toward the realization of effective AI research assistants. I will describethe activity, challenges, and progress of the different groups developing the key aspects of AuroraGPT. Iwill particularly focus on the task of conversational research assistant and discuss the evaluation of LLMs'scientific skills, their safety and trustworthiness, and the co-design of a scientific benchmark with domainexperts. |
| Description: | 2024 IEEE International Parallel and Distributed Processing Symposium (IPDPS).27-31 May 2024.San Francisco, CA, USA |
| Conference Name: | 2024 IEEE International Parallel and Distributed Processing Symposium (IPDPS) |
| Conference Place: | San Francisco, CA, USA |
| Rights: | © 2024 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. |
| Type: | proceedings (author version) |
| URI: | http://hdl.handle.net/2115/92940 |
| Appears in Collections: | 情報基盤センター (Information Initiative Center) > 雑誌発表論文等 (Peer-reviewed Journal Articles, etc)
|
Submitter: 深谷 猛
|
