Active constrained deep embedded clustering with dual source

Hazratgholizadeh, R.; Balafar, Mohammad Ali; Feizi‐Derakhshi, Mohammad‐Reza

doi:10.1007/s10489-022-03752-5

Cited by 4 publications

(3 citation statements)

References 50 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Lutz et al (2021) proposed an active clustering of training data in which the data is classified by the computer and the pairwise data are then sent to human experts for query. Hazratgholizadeh et al (2022) proposed an active constrained deep embedding clustering method that utilizes two parallel layers to select information and diversity constraints. Li et al (2021) proposed an adaptive criteria weights batch selection method that identifies the most informative pairs for semisupervised clustering through iterative means.…”

Section: Active Semi-supervised Clusteringmentioning

confidence: 99%

Deep Constrained Clustering with Active Learning

HUANG,

WEN,

DING

et al. 2023

SIC

View full text Add to dashboard Cite

Deep semi-supervised clustering approaches, which use supervised data to help the deep neural network acquire cluster-friendly representations, have improved clustering performance and simultaneously increased the semantic value of the clustering results. However, the majority of them cannot utilize both labeled and unlabeled data completely. Furthermore, in these methods, the supervised information is either passively acquired or randomly picked, which may be insufficient, redundant, and even decrease the performance of these models. This paper provides a deep semi-supervised clustering technique with active learning to address the problems mentioned above. The procedure is divided into two sections: model training and data labeling. In the model training section, the paired data is used to train the pseudo-Siamese network, and then the sub networks of the pseudo-Siamese network are fine-tuned using self-training. A new query strategy is devised in the data labeling part, which combines the traditional uncertainty query strategy with the deep Bayesian uncertainty query strategy. Finally, substantial tests are conducted to confirm the utility of the suggested approach on certain real-world data sets. The results of the tests demonstrate that both the suggested method and query strategy are practical.

show abstract

Section: Active Semi-supervised Clusteringmentioning

confidence: 99%

Deep Constrained Clustering with Active Learning

HUANG,

WEN,

DING

et al. 2023

SIC

View full text Add to dashboard Cite

show abstract

“…In the manual selection, the speckle noise inherent in SAR images will aggravate the difficulty of the identification of constraints. Some studies [44][45][46][47] have attempted to introduce remarkably effective constraint information, but their methods are only applicable in specific domains. Therefore, an active pairwise constraint learning method (APCL) is introduced to achieve reliable crop mapping from airborne SAR images via constrained time-series clustering.…”

Section: Study Area and Experimental Datamentioning

confidence: 99%

“…In the manual selection, the speckle noise inherent in SAR images will aggravate the difficulty of the identification of constraints. Some studies [44][45][46][47] have Remote Sens. 2022, 14, 6073 3 of 21 attempted to introduce remarkably effective constraint information, but their methods are only applicable in specific domains.…”

Section: Introductionmentioning

confidence: 99%

Active Pairwise Constraint Learning in Constrained Time-Series Clustering for Crop Mapping from Airborne SAR Imagery

et al. 2022

View full text Add to dashboard Cite

Airborne SAR is an important data source for crop mapping and has important applications in agricultural monitoring and food safety. However, the incidence-angle effects of airborne SAR imagery decrease the crop mapping accuracy. An active pairwise constraint learning method (APCL) is proposed for constrained time-series clustering to address this problem. APCL constructs two types of instance-level pairwise constraints based on the incidence angles of the samples and a non-iterative batch-mode active selection scheme: the must-link constraint, which links two objects of the same crop type with large differences in backscattering coefficients and the shapes of time-series curves; the cannot-link constraint, which links two objects of different crop types with only small differences in the values of backscattering coefficients. Experiments were conducted using 12 time-series images with incidence angles ranging from 21.2° to 64.3°, and the experimental results prove the effectiveness of APCL in improving crop mapping accuracy. More specifically, when using dynamic time warping (DTW) as the similarity measure, the kappa coefficient obtained by APCL was increased by 9.5%, 8.7%, and 5.2% compared to the results of the three other methods. It provides a new solution for reducing the incidence-angle effects in the crop mapping of airborne SAR time-series images.

show abstract