IEEE Geoscience and Remote Sensing Magazine - June 2016 - 14

0

0

1

1

(a)

(b)

FiGure 3. Comparing the performance of a multitask learning
method [46] for forest-cover estimation in four states in Brazil with
respect to a baseline approach of learning a single global model.
(a) The absolute residual errors of a single global model. (b) The
absolute residual errors of a multitask learning method [46]. The
color represents the average absolute error at a location across all
years, which ranges from 0 (no error) to 1 (maximum error).

vegetation types such as forests and shrublands. Multitask
learning, therefore, enforces a concept of coherence among
the model parameters of similar tasks, resulting in the
learning of physically meaningful models that adhere to
the domain understanding as
well as improve the detection
performance. The use of mula major challeNGe iN
titask learning frameworks
usiNG multitask learNiNG
for handling heterogeneity in
For chaNGe-detectioN
the data, even in the presence
problems is the Fact that
of limited training labels,
a majority oF multitask
would offer a global capability of change detection, in
learNiNG methods rely
contrast to existing changeoN a clear deFiNitioN oF
detection approaches that
the tasks aNd the relaare only applicable at local
tioNships amoNG them,
to regional scales due to the
Which is oFteN Not
heterogeneity in the data and
kNoWN explicitly.
limited availability of training instances.
In the context of using multitask learning frameworks for land-cover change detection,
it is often the case that the information about the composition of the tasks and the relationships between them is not
explicitly known because the knowledge about the partitioning of the overall data into homogeneous groups of instances is implicit in nature and guided by the specific application requirements. In such scenarios, information about
the exact number, distribution, and relationships among
the tasks can either be obtained using domain knowledge,
e.g., using information about the spatiotemporal autocorrrelation structure in the data or the hierarchical division of land-cover types using known taxonomies, or can
be discovered from the data using data-driven approaches.
As an example, a recent approach in [46] explored the use of
14

hierarchical clustering techniques for partitioning regions
into homogeneous groups of locations that belong to a
common vegetation type. This approach relied on the observation that locations belonging to the same vegetation
type show similar behavior in the temporal profile of their
vegetation across time, as demonstrated in [47]. Hence,
clustering locations on the basis of their contextual information contained in their vegetation time series over a long
period of time resulted in the discovery of groups of locations belonging to a common vegetation type. The learning
of a predictive model at every such cluster (i.e., vegetation
type) was then considered as a separate learning task. Furthermore, the hierarchical structure among the clusters,
which was found as a by-product of the clustering process,
provided a graph-based representation of the similarity
among the tasks, which was used for sharing the learning in a multitask learning framework.
Figure 3 shows the differences in the results of the multitask learning framework proposed in [46] and the traditional approach of learning a single, global predictive model
for the target application of estimating the amount of forest cover across four states in Brazil, i.e., Mato Grosso, Pará,
Amapá, and Roraima. The color at every pixel in Figure 3(a)
and (b) represents the magnitude of residual errors obtained
at that location by using the global model and the multitask
learning model, respectively, for estimating forest cover. It
can be seen that the method proposed in [46] provides significantly lower errors than the global model over a majority
of locations, illustrating the usefulness of multitask learning
methods for capturing the heterogeneity in the data, even
in the presence of a limited number of training labels. For
further examples, multitask learning has been explored for
remote sensing image classification in [48] and [49], while
the use of transfer learning methods for land-cover change
detection has been explored in [34], [50], and [51], which
use a similar formulation as multitask learning methods.
A major challenge in using multitask learning for changedetection problems is the fact that a majority of multitask
learning methods rely on a clear definition of the tasks and
the relationships among them, which is often not known
explicitly. For example, since every atomic task represents
the learning of a change-detection model at a homogeneous
group of instances, the notion of a task depends heavily on
the choice of the contextual variables used for representing
the heterogeneity in the data, and the use of the clustering
strategy and is not well defined. Hence there is a need to
develop novel multitask learning frameworks that can first
infer the heterogeneity in the data by partitioning the data
into homogeneous groups and learning the relationships
among the groups, and then utilize this information for
learning a change-detection model at every such group.
MULTI-INSTANCE LEARNING
In the classical setting, the basic unit of classification
is a data instance that comprises a single observation
of features that has an associated label. However, in
ieee Geoscience and remote sensing magazine

june 2016



Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - June 2016

IEEE Geoscience and Remote Sensing Magazine - June 2016 - Cover1
IEEE Geoscience and Remote Sensing Magazine - June 2016 - Cover2
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 1
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 2
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 3
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 4
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 5
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 6
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 7
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 8
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 9
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 10
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 11
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 12
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 13
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 14
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 15
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 16
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 17
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 18
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 19
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 20
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 21
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 22
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 23
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 24
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 25
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 26
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 27
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 28
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 29
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 30
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 31
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 32
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 33
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 34
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 35
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 36
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 37
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 38
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 39
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 40
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 41
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 42
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 43
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 44
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 45
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 46
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 47
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 48
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 49
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 50
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 51
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 52
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 53
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 54
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 55
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 56
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 57
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 58
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 59
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 60
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 61
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 62
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 63
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 64
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 65
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 66
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 67
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 68
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 69
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 70
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 71
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 72
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 73
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 74
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 75
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 76
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 77
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 78
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 79
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 80
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 81
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 82
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 83
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 84
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 85
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 86
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 87
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 88
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 89
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 90
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 91
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 92
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 93
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 94
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 95
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 96
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 97
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 98
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 99
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 100
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 101
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 102
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 103
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 104
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 105
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 106
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 107
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 108
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 109
IEEE Geoscience and Remote Sensing Magazine - June 2016 - 110
IEEE Geoscience and Remote Sensing Magazine - June 2016 - Cover3
IEEE Geoscience and Remote Sensing Magazine - June 2016 - Cover4
https://www.nxtbook.com/nxtbooks/ieee/geoscience_december2023
https://www.nxtbook.com/nxtbooks/ieee/geoscience_september2023
https://www.nxtbook.com/nxtbooks/ieee/geoscience_june2023
https://www.nxtbook.com/nxtbooks/ieee/geoscience_march2023
https://www.nxtbook.com/nxtbooks/ieee/geoscience_december2022
https://www.nxtbook.com/nxtbooks/ieee/geoscience_september2022
https://www.nxtbook.com/nxtbooks/ieee/geoscience_june2022
https://www.nxtbook.com/nxtbooks/ieee/geoscience_march2022
https://www.nxtbook.com/nxtbooks/ieee/geoscience_december2021
https://www.nxtbook.com/nxtbooks/ieee/geoscience_september2021
https://www.nxtbook.com/nxtbooks/ieee/geoscience_june2021
https://www.nxtbook.com/nxtbooks/ieee/geoscience_march2021
https://www.nxtbook.com/nxtbooks/ieee/geoscience_december2020
https://www.nxtbook.com/nxtbooks/ieee/geoscience_september2020
https://www.nxtbook.com/nxtbooks/ieee/geoscience_june2020
https://www.nxtbook.com/nxtbooks/ieee/geoscience_march2020
https://www.nxtbook.com/nxtbooks/ieee/geoscience_december2019
https://www.nxtbook.com/nxtbooks/ieee/geoscience_september2019
https://www.nxtbook.com/nxtbooks/ieee/geoscience_june2019
https://www.nxtbook.com/nxtbooks/ieee/geoscience_march2019
https://www.nxtbook.com/nxtbooks/ieee/geoscience_december2018
https://www.nxtbook.com/nxtbooks/ieee/geoscience_september2018
https://www.nxtbook.com/nxtbooks/ieee/geoscience_june2018
https://www.nxtbook.com/nxtbooks/ieee/geoscience_march2018
https://www.nxtbook.com/nxtbooks/ieee/geoscience_december2017
https://www.nxtbook.com/nxtbooks/ieee/geoscience_september2017
https://www.nxtbook.com/nxtbooks/ieee/geoscience_june2017
https://www.nxtbook.com/nxtbooks/ieee/geoscience_march2017
https://www.nxtbook.com/nxtbooks/ieee/geoscience_december2016
https://www.nxtbook.com/nxtbooks/ieee/geoscience_september2016
https://www.nxtbook.com/nxtbooks/ieee/geoscience_june2016
https://www.nxtbook.com/nxtbooks/ieee/geoscience_march2016
https://www.nxtbook.com/nxtbooks/ieee/geoscience_december2015
https://www.nxtbook.com/nxtbooks/ieee/geoscience_september2015
https://www.nxtbook.com/nxtbooks/ieee/geoscience_june2015
https://www.nxtbook.com/nxtbooks/ieee/geoscience_march2015
https://www.nxtbook.com/nxtbooks/ieee/geoscience_december2014
https://www.nxtbook.com/nxtbooks/ieee/geoscience_september2014
https://www.nxtbook.com/nxtbooks/ieee/geoscience_june2014
https://www.nxtbook.com/nxtbooks/ieee/geoscience_march2014
https://www.nxtbook.com/nxtbooks/ieee/geoscience_december2013
https://www.nxtbook.com/nxtbooks/ieee/geoscience_september2013
https://www.nxtbook.com/nxtbooks/ieee/geoscience_june2013
https://www.nxtbook.com/nxtbooks/ieee/geoscience_march2013
https://www.nxtbookmedia.com