text
stringlengths
0
820
Class 4
Class 5
Class 6
Class 7Class 8
Class 9
Class 10
Class 11
Class 12
Class 13
Class 14
Class 1
Class 2
Class 3
Class 4
Class 5
Class 6
Class 7Class 8
Class 9
Class 10
Class 11
Class 12
Class 13
Class 14
01,0002,0003,0004,0005,0006,0007,000
Source Domain
Target Domain (b)
05 0 100 150
(c) (a)
Figu Re 4. (a) A false color composition of a portion of the hyperspectral data set. (b) The mean spectral signature of the classes on the
source domain. (c) The mean spectral signature of the classes on the target domain.
in tHe Remote SenSing
LiteRatuRe, SeVeRaL
tecHniQueS Ha Ve Been
pReSente D to S oLVe tHe
tRanSFeR-LeaRning
pRoBLem iRReSpecti Ve
oF tHe cau Se oF tHe Data
Set SHiFt Between
SouRce an D taRget
DomainS.
Authorized licensed use limited to: ASU Library. Downloaded on March 08,2024 at 03:13:37 UTC from IEEE Xplore. Restrictions apply.
june 2016 ieee Geoscience and remote sensin G ma Gazine 47
In the following, we report the experimental results ob -
tained on a hyperspectral image acquired by the Hyperion
sensor of the Earth Observation 1 satellite in an area of the
Okavango Delta, Botswana [ 32] (see Figure 4 ). For more in -
formation about the experimental setting and the obtained
results, see [ 30]. The labeled reference samples were col -
lected on two spatially disjoint areas with slightly different
characteristics, thus representing two different domains.
The samples taken on the first area, which was considered
as the source domain, were partitioned into a training set
Ts and a test set Ts by random sampling. Samples taken on
the second area (i.e., the target domain) were used to derive
a training set Tt and test set Tt according to the same pro -
cedure. The estimated Pareto front for the selection of six features is reported in Figure 5 .
Each point on the two graphs corresponds to a different
selected feature subset
F, i.e, a feature set minimizing (1 ).
In Figure 5(a), the color of the points indicates the overall accuracy (OA) obtained on the source-domain test set
Ts
using an SVM classifier trained using Ts (according to the
reported color scale bar). In Figure 5(b), the color indicates the OA obtained by the same SVM classifier on the target-
domain test set
.Tt
The results show that the solutions with higher relevance
D result in better classification accuracies on the source do -
main. However, relevance alone is not sufficient for select -
ing features that are stable for the classification on the target domain. We observe that the most accurate solutions on the target domain
Tt are those that exhibit a good tradeoff be -
tween the relevance and invariance terms, which confirms
the importance of the invariance term and shows that the
P measure is able to capture the information of feature
stability. To select the subset of features that leads to good generalization capabilities on different domains, tradeoff
solutions between the two competing objectives should be
identified. The selected subset of features results in an OA
of 91% on the source domain and 80.7% on the target. The
set of features selected according to the optimization of
D
results in an OA of 92.7% on the source but only 64.4% on the target, showing that the features selected by accounting for the data set shift between the domains can significantly improve the generalization capability on the target domain.
ADAPTING DATA DISTRIBUTIONS
The second family reviewed considers DA methods that
aim to adapt the representation of the original data, regard -
less of the model that will process them afterward. A review
of the methods proposed in computer vision and machine
learning can be found in [ 16]. Here, we will focus on the ap -
proaches proposed in the remote sensing literature. This type
of adaptation is often done by relative normalization methods,
i.e., methods that do not provide physical units as an output
but that instead provide similarly distributed digital numbers.
Their aim is to make the data
distri butions more similar
across the domains to train a
single model that can simul -
taneously classify the source
and target domains.
In general, a data represen -
tation transformation with the aim of making data sources more similar should have the
following desirable properties:
◗The method should be able