Periodic activation function for fast on-line EKF training and pruning

Sheng-Jiang Chang; Kwok-Wo Wong; Chi-Sing Leung

doi:10.1049/el:19981554

Periodic activation function for fast on-line EKF training and pruning

Sheng-Jiang Chang
, Kwok-Wo Wong
, Chi-Sing Leung

Department of Electrical Engineering

Research output: Journal Publications and Reviews › RGC 21 - Publication in refereed journal › peer-review

7 Citations (Scopus)

Abstract

In the use of the local extended Kalman filter algorithm for network training and pruning, the authors propose replacing the monotonical increasing sigmoid activation function with a periodic function. Computer simulations show that the rate of convergence is substantially improved and that more insignificant weights can be pruned.

Original language	English
Pages (from-to)	2255-2257
Journal	Electronics Letters
Volume	34
Issue number	23
DOIs	https://doi.org/10.1049/el:19981554
Publication status	Published - 12 Nov 1998

Access Output

10.1049/el:19981554

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

@article{272cd799d47b4a1598b0242fcc0573d5,

title = "Periodic activation function for fast on-line EKF training and pruning",

abstract = "In the use of the local extended Kalman filter algorithm for network training and pruning, the authors propose replacing the monotonical increasing sigmoid activation function with a periodic function. Computer simulations show that the rate of convergence is substantially improved and that more insignificant weights can be pruned.",

author = "Sheng-Jiang Chang and Kwok-Wo Wong and Chi-Sing Leung",

year = "1998",

month = nov,

day = "12",

doi = "10.1049/el:19981554",

language = "English",

volume = "34",

pages = "2255--2257",

journal = "Electronics Letters",

issn = "0013-5194",

publisher = "John Wiley \& Sons",

number = "23",

}

TY - JOUR

T1 - Periodic activation function for fast on-line EKF training and pruning

AU - Chang, Sheng-Jiang

AU - Wong, Kwok-Wo

AU - Leung, Chi-Sing

PY - 1998/11/12

Y1 - 1998/11/12

N2 - In the use of the local extended Kalman filter algorithm for network training and pruning, the authors propose replacing the monotonical increasing sigmoid activation function with a periodic function. Computer simulations show that the rate of convergence is substantially improved and that more insignificant weights can be pruned.

AB - In the use of the local extended Kalman filter algorithm for network training and pruning, the authors propose replacing the monotonical increasing sigmoid activation function with a periodic function. Computer simulations show that the rate of convergence is substantially improved and that more insignificant weights can be pruned.

UR - https://www.scopus.com/pages/publications/0032206420

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-0032206420&origin=recordpage

U2 - 10.1049/el:19981554

DO - 10.1049/el:19981554

M3 - RGC 21 - Publication in refereed journal

SN - 0013-5194

VL - 34

SP - 2255

EP - 2257

JO - Electronics Letters

JF - Electronics Letters

IS - 23

ER -

Periodic activation function for fast on-line EKF training and pruning

Abstract

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Cite this