Performance of the Bayesian online algorithm for the perceptron

Evaldo Araújo de Oliveira; Roberto C. Alamino

doi:10.1109/TNN.2007.891189

Performance of the Bayesian online algorithm for the perceptron

Evaldo Araújo de Oliveira
, Roberto C. Alamino

College of Engineering and Physical Sciences

Universidade de São Paulo

Research output: Contribution to journal › Letter, comment/opinion or interview › peer-review

5 Citations (Scopus)

Abstract

In this letter, we derive continuum equations for the generalization error of the Bayesian online algorithm (BOnA) for the one-layer perceptron with a spherical covariance matrix using the Rosenblatt potential and show, by numerical calculations, that the asymptotic performance of the algorithm is the same as the one for the optimal algorithm found by means of variational methods with the added advantage that the BOnA does not use any inaccessible information during learning.

Original language	English
Pages (from-to)	902-905
Number of pages	4
Journal	IEEE Transactions on Neural Networks and Learning Systems
Volume	18
Issue number	3
DOIs	https://doi.org/10.1109/TNN.2007.891189
Publication status	Published - May 2007

Keywords

Bayesian algorithms
online gradient methods
pattern classification

Access to Document

10.1109/TNN.2007.891189

Cite this

@article{82d67a10e0d34dc9b6f54a626e5f29d7,

title = "Performance of the Bayesian online algorithm for the perceptron",

abstract = "In this letter, we derive continuum equations for the generalization error of the Bayesian online algorithm (BOnA) for the one-layer perceptron with a spherical covariance matrix using the Rosenblatt potential and show, by numerical calculations, that the asymptotic performance of the algorithm is the same as the one for the optimal algorithm found by means of variational methods with the added advantage that the BOnA does not use any inaccessible information during learning.",

keywords = "Bayesian algorithms, online gradient methods, pattern classification",

author = "\{de Oliveira\}, \{Evaldo Ara{\'u}jo\} and Alamino, \{Roberto C.\}",

year = "2007",

month = may,

doi = "10.1109/TNN.2007.891189",

language = "English",

volume = "18",

pages = "902--905",

journal = "IEEE Transactions on Neural Networks and Learning Systems",

issn = "1045-9227",

publisher = "IEEE",

number = "3",

}

TY - JOUR

T1 - Performance of the Bayesian online algorithm for the perceptron

AU - de Oliveira, Evaldo Araújo

AU - Alamino, Roberto C.

PY - 2007/5

Y1 - 2007/5

N2 - In this letter, we derive continuum equations for the generalization error of the Bayesian online algorithm (BOnA) for the one-layer perceptron with a spherical covariance matrix using the Rosenblatt potential and show, by numerical calculations, that the asymptotic performance of the algorithm is the same as the one for the optimal algorithm found by means of variational methods with the added advantage that the BOnA does not use any inaccessible information during learning.

AB - In this letter, we derive continuum equations for the generalization error of the Bayesian online algorithm (BOnA) for the one-layer perceptron with a spherical covariance matrix using the Rosenblatt potential and show, by numerical calculations, that the asymptotic performance of the algorithm is the same as the one for the optimal algorithm found by means of variational methods with the added advantage that the BOnA does not use any inaccessible information during learning.

KW - Bayesian algorithms

KW - online gradient methods

KW - pattern classification

UR - http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=4182376

UR - https://www.scopus.com/pages/publications/34248663836

U2 - 10.1109/TNN.2007.891189

DO - 10.1109/TNN.2007.891189

M3 - Letter, comment/opinion or interview

C2 - 17526354

AN - SCOPUS:34248663836

SN - 1045-9227

VL - 18

SP - 902

EP - 905

JO - IEEE Transactions on Neural Networks and Learning Systems

JF - IEEE Transactions on Neural Networks and Learning Systems

IS - 3

ER -

Performance of the Bayesian online algorithm for the perceptron

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this