國立虎尾科技大學 |

Data Attribution : = From Classifiers to Generative Models.

紀錄類型:	書目-語言資料,手稿 : Monograph/item
正題名/作者:	Data Attribution :/
其他題名:	From Classifiers to Generative Models.
作者:	Georgiev, Kristian.
面頁冊數:	1 online resource (132 pages)
附註:	Source: Masters Abstracts International, Volume: 85-09.
Contained By:	Masters Abstracts International85-09.
標題:	Computer engineering. -
電子資源:	click for full text (PQDT)
ISBN:	9798381958911

Data Attribution : = From Classifiers to Generative Models.
Georgiev, Kristian.

Data Attribution :From Classifiers to Generative Models. - 1 online resource (132 pages)

Source: Masters Abstracts International, Volume: 85-09.

Thesis (M.S.)--Massachusetts Institute of Technology, 2023.

Includes bibliographical references

The goal of data attribution is to trace model predictions back to training data. Despite a long line of work towards this goal, existing approaches to data attribution tend to force users to choose between computational tractability and efficacy. That is, computationally tractable methods can struggle with accurately attributing model predictions in non-convex settings (e.g., in the context of deep neural networks), while methods that are effective in such regimes require training thousands of models, which makes them impractical for large models or datasets. Moreover, existing methods are often tailored to the supervised learning setting, and are not well-defined for generative models.In this thesis, we introduce TRAK (Tracing with the Randomly-projected After Kernel), a data attribution method that is both effective and computationally tractable for large-scale, differentiable models. In particular, by leveraging only a handful of trained models, TRAK can match the performance of attribution methods that require training thousands of models. We first demonstrate the utility of TRAK across various modalities and scales in the supervised setting: image classifiers trained on ImageNet, vision-language models (CLIP), and language models (BERT and mT5). Then, we extend TRAK to the generative setting, and show that it can be used to attribute different classes of diffusion models (DDPMs and LDMs).

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2024

Mode of access: World Wide Web

ISBN: 9798381958911Subjects--Topical Terms:

569006
Computer engineering.
Subjects--Index Terms:

Training dataIndex Terms--Genre/Form:

554714
Electronic books.

Data Attribution : = From Classifiers to Generative Models.
LDR:02756ntm a22003857 4500 001 1146482
005 20240812064627.5
006 m o d
007 cr bn ---uuuuu
008 250605s2023 xx obm 000 0 eng d
020 $a 9798381958911
035 $a (MiAaPQ)AAI31091687
035 $a (MiAaPQ)MIT1721_1_152676
035 $a AAI31091687
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Georgiev, Kristian. $3 1471875
245 1 0 $a Data Attribution : $b From Classifiers to Generative Models.
264 0 $c 2023
300 $a 1 online resource (132 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: Masters Abstracts International, Volume: 85-09.
500 $a Advisor: Madry, Aleksander.
502 $a Thesis (M.S.)--Massachusetts Institute of Technology, 2023.
504 $a Includes bibliographical references
520 $a The goal of data attribution is to trace model predictions back to training data. Despite a long line of work towards this goal, existing approaches to data attribution tend to force users to choose between computational tractability and efficacy. That is, computationally tractable methods can struggle with accurately attributing model predictions in non-convex settings (e.g., in the context of deep neural networks), while methods that are effective in such regimes require training thousands of models, which makes them impractical for large models or datasets. Moreover, existing methods are often tailored to the supervised learning setting, and are not well-defined for generative models.In this thesis, we introduce TRAK (Tracing with the Randomly-projected After Kernel), a data attribution method that is both effective and computationally tractable for large-scale, differentiable models. In particular, by leveraging only a handful of trained models, TRAK can match the performance of attribution methods that require training thousands of models. We first demonstrate the utility of TRAK across various modalities and scales in the supervised setting: image classifiers trained on ImageNet, vision-language models (CLIP), and language models (BERT and mT5). Then, we extend TRAK to the generative setting, and show that it can be used to attribute different classes of diffusion models (DDPMs and LDMs).
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2024
538 $a Mode of access: World Wide Web
650 4 $a Computer engineering. $3 569006
650 4 $a Electrical engineering. $3 596380
653 $a Training data
653 $a Vision-language models
653 $a Generative models
653 $a Data attribution
655 7 $a Electronic books. $2 local $3 554714
690 $a 0544
690 $a 0464
710 2 $a ProQuest Information and Learning Co. $3 1178819
710 2 $a Massachusetts Institute of Technology. $b Department of Electrical Engineering and Computer Science. $3 1467552
773 0 $t Masters Abstracts International $g 85-09.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31091687 $z click for full text (PQDT)