Title SVM active learning through significance space construction Author-s Contact lnfo Department Advanced Lab for Intelligent Systems Research (ALISR) Major Information and Communication Technologies (ICT) citation Year of

munchsistersΤεχνίτη Νοημοσύνη και Ρομποτική

17 Οκτ 2013 (πριν από 4 χρόνια και 27 μέρες)

91 εμφανίσεις



Title

SVM active learning through significance space construction

Author
-
s

Pasolli, E.


Melgani, F.


Bazi, Y.


Contact lnfo

ybazi@ksu.edu.sa

Department

Advanced Lab for Intelligent Systems Research (ALISR)

Major

Information and Communication
Technologies (ICT)

citation

Geoscience and Remote Sensing Letters, IEEE
, Volume: 8
Issue: 3


On page
(s): 431
-

435

Year of
Publication

201
0

Publisher

Geoscience and Remote Sensing Letters, IEEE

Sponsor

IEEE Geoscience

and Remote Sensing Society

Type of
Publication

Journal paper

ISSN

1545
-
598X

URI/DOI


10.1109/LGRS.2010.2083630

http://ieeexplore.ieee.org/search/freesrchabstract.jsp?tp=&arnumber=56
28257&q
ueryText%3DBazi%26openedRefinements%3D*%26searchField%3DSearch+All

Full Text
(Yes,No)

Yes

Key words

Active learning, hyperspectral images, support vector machines (SVMs), very
-
high
-
resolution (VHR) images

Abstract

Active learning is showing to
be a useful approach to improve the efficiency of
the classification process for remote sensing images. This letter introduces a new
active learning strategy specifically developed for support vector machine (SVM)
classification. It relies on the idea of t
he following: 1) reformulating the original
classification problem into a new problem where it is needed to discriminate
between significant and non significant samples, according to a concept of
significance which is proper to the SVM theory; and 2) const
ructing the
corresponding significance space to suitably guide the selection of the samples
potentially useful to better deal with the original classification problem.
Experiments were conducted on both multi
-

and hyperspectral images. Results
show interes
ting advantages of the proposed method in terms of convergence
speed, stability, and sparseness.