volume 26 issue 2 pages 289-315

On Early Stopping in Gradient Descent Learning

Publication typeJournal Article
Publication date2007-04-04
scimago Q1
wos Q1
SJR1.871
CiteScore5.0
Impact factor1.2
ISSN01764276, 14320940
General Mathematics
Computational Mathematics
Analysis
Abstract
In this paper we study a family of gradient descent algorithms to approximate the regression function from reproducing kernel Hilbert spaces (RKHSs), the family being characterized by a polynomial decreasing rate of step sizes (or learning rate). By solving a bias-variance trade-off we obtain an early stopping rule and some probabilistic upper bounds for the convergence of the algorithms. We also discuss the implication of these results in the context of classification where some fast convergence rates can be achieved for plug-in classifiers. Some connections are addressed with Boosting, Landweber iterations, and the online learning algorithms as stochastic approximations of the gradient descent method.
Found 

Top-30

Journals

2
4
6
8
10
12
14
Lecture Notes in Computer Science
14 publications, 2.18%
IEEE Access
14 publications, 2.18%
Applied Sciences (Switzerland)
10 publications, 1.56%
Remote Sensing
8 publications, 1.24%
Scientific Reports
8 publications, 1.24%
Applied and Computational Harmonic Analysis
8 publications, 1.24%
Mathematics
7 publications, 1.09%
IEEE Transactions on Neural Networks and Learning Systems
7 publications, 1.09%
Analysis and Applications
6 publications, 0.93%
Electronics (Switzerland)
5 publications, 0.78%
Sensors
5 publications, 0.78%
Multimedia Tools and Applications
5 publications, 0.78%
Neural Networks
5 publications, 0.78%
Inverse Problems
5 publications, 0.78%
Lecture Notes in Networks and Systems
5 publications, 0.78%
Constructive Approximation
4 publications, 0.62%
Mechanical Systems and Signal Processing
4 publications, 0.62%
Neurocomputing
4 publications, 0.62%
Applied Energy
4 publications, 0.62%
Applied Soft Computing Journal
4 publications, 0.62%
Communications in Computer and Information Science
4 publications, 0.62%
Physical Review D
3 publications, 0.47%
Neural Computation
3 publications, 0.47%
IEEE Transactions on Geoscience and Remote Sensing
3 publications, 0.47%
SIAM Journal on Numerical Analysis
3 publications, 0.47%
Annals of Statistics
3 publications, 0.47%
Entropy
3 publications, 0.47%
Neural Processing Letters
3 publications, 0.47%
Applied Intelligence
3 publications, 0.47%
2
4
6
8
10
12
14

Publishers

20
40
60
80
100
120
140
Institute of Electrical and Electronics Engineers (IEEE)
137 publications, 21.31%
Springer Nature
119 publications, 18.51%
Elsevier
114 publications, 17.73%
MDPI
64 publications, 9.95%
Association for Computing Machinery (ACM)
24 publications, 3.73%
Wiley
19 publications, 2.95%
Oxford University Press
12 publications, 1.87%
World Scientific
10 publications, 1.56%
Cold Spring Harbor Laboratory
10 publications, 1.56%
American Physical Society (APS)
8 publications, 1.24%
IOP Publishing
8 publications, 1.24%
American Chemical Society (ACS)
8 publications, 1.24%
Taylor & Francis
8 publications, 1.24%
Institute of Mathematical Statistics
7 publications, 1.09%
Frontiers Media S.A.
7 publications, 1.09%
Copernicus
6 publications, 0.93%
AIP Publishing
5 publications, 0.78%
Hindawi Limited
5 publications, 0.78%
SAGE
4 publications, 0.62%
MIT Press
4 publications, 0.62%
Society for Industrial and Applied Mathematics (SIAM)
3 publications, 0.47%
Public Library of Science (PLoS)
3 publications, 0.47%
Walter de Gruyter
3 publications, 0.47%
SPIE-Intl Soc Optical Eng
2 publications, 0.31%
Institute for Operations Research and the Management Sciences (INFORMS)
2 publications, 0.31%
Tech Science Press
2 publications, 0.31%
Social Science Electronic Publishing
2 publications, 0.31%
Royal Society of Chemistry (RSC)
2 publications, 0.31%
Society of Exploration Geophysicists
2 publications, 0.31%
20
40
60
80
100
120
140
  • We do not take into account publications without a DOI.
  • Statistics recalculated weekly.

Are you a researcher?

Create a profile to get free access to personal recommendations for colleagues and new articles.
Metrics
643
Share
Cite this
GOST |
Cite this
GOST Copy
Yao Y. et al. On Early Stopping in Gradient Descent Learning // Constructive Approximation. 2007. Vol. 26. No. 2. pp. 289-315.
GOST all authors (up to 50) Copy
Yao Y., Rosasco L., CAPONNETTO A. On Early Stopping in Gradient Descent Learning // Constructive Approximation. 2007. Vol. 26. No. 2. pp. 289-315.
RIS |
Cite this
RIS Copy
TY - JOUR
DO - 10.1007/s00365-006-0663-2
UR - https://doi.org/10.1007/s00365-006-0663-2
TI - On Early Stopping in Gradient Descent Learning
T2 - Constructive Approximation
AU - Yao, Yuan
AU - Rosasco, Lorenzo
AU - CAPONNETTO, ANDREA
PY - 2007
DA - 2007/04/04
PB - Springer Nature
SP - 289-315
IS - 2
VL - 26
SN - 0176-4276
SN - 1432-0940
ER -
BibTex |
Cite this
BibTex (up to 50 authors) Copy
@article{2007_Yao,
author = {Yuan Yao and Lorenzo Rosasco and ANDREA CAPONNETTO},
title = {On Early Stopping in Gradient Descent Learning},
journal = {Constructive Approximation},
year = {2007},
volume = {26},
publisher = {Springer Nature},
month = {apr},
url = {https://doi.org/10.1007/s00365-006-0663-2},
number = {2},
pages = {289--315},
doi = {10.1007/s00365-006-0663-2}
}
MLA
Cite this
MLA Copy
Yao, Yuan, et al. “On Early Stopping in Gradient Descent Learning.” Constructive Approximation, vol. 26, no. 2, Apr. 2007, pp. 289-315. https://doi.org/10.1007/s00365-006-0663-2.