Generalization Bounds for Deep Transfer Learning Using Majority Predictor Accuracy

Cuong N. Nguyen; Lam Si Tung Ho; Vu Dinh; Tal Hassner; Cuong V. Nguyen

Generalization Bounds for Deep Transfer Learning Using Majority Predictor Accuracy

Cuong N. Nguyen, Lam Si Tung Ho, Vu Dinh, Tal Hassner, Cuong V. Nguyen

פרסום מחקרי: פרק בספר / בדוח / בכנס › פרסום בספר כנס › ביקורת עמיתים

תקציר

We analyze new generalization bounds for deep learning models trained by transfer learning from a source to a target task. Our bounds utilize a quantity called the majority predictor accuracy, which can be computed efficiently from data. We show that our theory is useful in practice since it implies that the majority predictor accuracy can be used as a transferability measure, a fact that is also validated by our experiments.

שפה מקורית	אנגלית
כותר פרסום המארח	Proceedings of 2022 International Symposium on Information Theory and Its Applications, ISITA 2022
מוציא לאור	Institute of Electrical and Electronics Engineers Inc.
עמודים	139-143
מספר עמודים	5
מסת"ב (אלקטרוני)	9784885523410
סטטוס פרסום	פורסם - 2022
פורסם באופן חיצוני	כן
אירוע	17th International Symposium on Information Theory and Its Applications, ISITA 2022 - Tsukuba, Ibaraki, יפן משך הזמן: 17 אוק׳ 2022 → 19 אוק׳ 2022

סדרות פרסומים

שם	Proceedings of 2022 International Symposium on Information Theory and Its Applications, ISITA 2022

כנס

כנס	17th International Symposium on Information Theory and Its Applications, ISITA 2022
מדינה/אזור	יפן
עיר	Tsukuba, Ibaraki
תקופה	17/10/22 → 19/10/22

הערה ביבליוגרפית

Publisher Copyright:
© 2022 IEICE.

קבצים וקישורים אחרים

Link to publication in Scopus

פורמט ציטוט ביבליוגרפי

Nguyen, C. N., Ho, L. S. T., Dinh, V., Hassner, T., & Nguyen, C. V. (2022). Generalization Bounds for Deep Transfer Learning Using Majority Predictor Accuracy. ב- Proceedings of 2022 International Symposium on Information Theory and Its Applications, ISITA 2022 (עמודים 139-143). (Proceedings of 2022 International Symposium on Information Theory and Its Applications, ISITA 2022). Institute of Electrical and Electronics Engineers Inc..

Nguyen, Cuong N. ; Ho, Lam Si Tung ; Dinh, Vu et al. / Generalization Bounds for Deep Transfer Learning Using Majority Predictor Accuracy. Proceedings of 2022 International Symposium on Information Theory and Its Applications, ISITA 2022. Institute of Electrical and Electronics Engineers Inc., 2022. עמודים 139-143 (Proceedings of 2022 International Symposium on Information Theory and Its Applications, ISITA 2022).

@inproceedings{8e05eec0b7d14b909bbbf4732b5adcfc,

title = "Generalization Bounds for Deep Transfer Learning Using Majority Predictor Accuracy",

abstract = "We analyze new generalization bounds for deep learning models trained by transfer learning from a source to a target task. Our bounds utilize a quantity called the majority predictor accuracy, which can be computed efficiently from data. We show that our theory is useful in practice since it implies that the majority predictor accuracy can be used as a transferability measure, a fact that is also validated by our experiments.",

author = "Nguyen, {Cuong N.} and Ho, {Lam Si Tung} and Vu Dinh and Tal Hassner and Nguyen, {Cuong V.}",

note = "Publisher Copyright: {\textcopyright} 2022 IEICE.; 17th International Symposium on Information Theory and Its Applications, ISITA 2022 ; Conference date: 17-10-2022 Through 19-10-2022",

year = "2022",

language = "אנגלית",

series = "Proceedings of 2022 International Symposium on Information Theory and Its Applications, ISITA 2022",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "139--143",

booktitle = "Proceedings of 2022 International Symposium on Information Theory and Its Applications, ISITA 2022",

address = "ארצות הברית",

}

Nguyen, CN, Ho, LST, Dinh, V, Hassner, T & Nguyen, CV 2022, Generalization Bounds for Deep Transfer Learning Using Majority Predictor Accuracy. ב- Proceedings of 2022 International Symposium on Information Theory and Its Applications, ISITA 2022. Proceedings of 2022 International Symposium on Information Theory and Its Applications, ISITA 2022, Institute of Electrical and Electronics Engineers Inc., עמודים 139-143, 17th International Symposium on Information Theory and Its Applications, ISITA 2022, Tsukuba, Ibaraki, יפן, 17/10/22.

Generalization Bounds for Deep Transfer Learning Using Majority Predictor Accuracy. / Nguyen, Cuong N.; Ho, Lam Si Tung; Dinh, Vu et al.
Proceedings of 2022 International Symposium on Information Theory and Its Applications, ISITA 2022. Institute of Electrical and Electronics Engineers Inc., 2022. עמוד 139-143 (Proceedings of 2022 International Symposium on Information Theory and Its Applications, ISITA 2022).

פרסום מחקרי: פרק בספר / בדוח / בכנס › פרסום בספר כנס › ביקורת עמיתים

TY - GEN

T1 - Generalization Bounds for Deep Transfer Learning Using Majority Predictor Accuracy

AU - Nguyen, Cuong N.

AU - Ho, Lam Si Tung

AU - Dinh, Vu

AU - Hassner, Tal

AU - Nguyen, Cuong V.

PY - 2022

Y1 - 2022

N2 - We analyze new generalization bounds for deep learning models trained by transfer learning from a source to a target task. Our bounds utilize a quantity called the majority predictor accuracy, which can be computed efficiently from data. We show that our theory is useful in practice since it implies that the majority predictor accuracy can be used as a transferability measure, a fact that is also validated by our experiments.

AB - We analyze new generalization bounds for deep learning models trained by transfer learning from a source to a target task. Our bounds utilize a quantity called the majority predictor accuracy, which can be computed efficiently from data. We show that our theory is useful in practice since it implies that the majority predictor accuracy can be used as a transferability measure, a fact that is also validated by our experiments.

UR - http://www.scopus.com/inward/record.url?scp=85170075675&partnerID=8YFLogxK

M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???

AN - SCOPUS:85170075675

T3 - Proceedings of 2022 International Symposium on Information Theory and Its Applications, ISITA 2022

SP - 139

EP - 143

BT - Proceedings of 2022 International Symposium on Information Theory and Its Applications, ISITA 2022

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 17th International Symposium on Information Theory and Its Applications, ISITA 2022

Y2 - 17 October 2022 through 19 October 2022

ER -

Nguyen CN, Ho LST, Dinh V, Hassner T, Nguyen CV. Generalization Bounds for Deep Transfer Learning Using Majority Predictor Accuracy. ב- Proceedings of 2022 International Symposium on Information Theory and Its Applications, ISITA 2022. Institute of Electrical and Electronics Engineers Inc. 2022. עמוד 139-143. (Proceedings of 2022 International Symposium on Information Theory and Its Applications, ISITA 2022).

Generalization Bounds for Deep Transfer Learning Using Majority Predictor Accuracy

תקציר

סדרות פרסומים

כנס

הערה ביבליוגרפית

קבצים וקישורים אחרים

טביעת אצבע

פורמט ציטוט ביבליוגרפי