Active Structure Learning for Block Diagonal Subspace Clustering

Xie, Ziqi; Wang, Lihong

doi:10.1109/cisp-bmei51763.2020.9263491

Cited by 1 publication

(4 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…(2) Z M ∪C = 0 [16], [29]. Set 0 at all constrained entries in the representation matrix, and then reconstruct the affinity matrix W for the spectral clustering, i.e., W ij is set to v 0 , if x i and x j are must-linked, and 0 if they are cannotlinked.…”

Section: Discussion On Constraint Methodsmentioning

confidence: 99%

“…al. [29] proposed a constrained SBDR subspace clustering algorithm. The objective function of SBDR is re-optimized under an additional semisupervised constraint, in which the representative coefficients between both must-linked and cannot-linked nodes are set to 0 to clean the noise in the representation matrix.…”

Section: A Bdrmentioning

confidence: 99%

“…The spatial information is included through a joint sparsity constraint on the coefficient matrix of each local region and the available label information is incorporated by zeroing the entries of the sparse coefficient matrix corresponding to the data points from different classes. To the best of our knowledge, our previous work [29] is the first one on constrained BDR. Specifically, this work incorporates pairwise constraints into the structured subspace clustering with block diagonal representation (SBDR).…”

Section: B Constrained Subspace Clusteringmentioning

confidence: 99%

“…In 2014, Biswas et al [36] used Jac-card's coefficient (JCC) as a measure of similarity between two clustering results, and selected the pair of nodes which produced the maximal expected change in the clustering. Xie et al [29] proposed an active learning strategy for block diagonal subspace clustering, which iteratively selects a point inside each block as a representative point of the block and a point outside the block as a suspected missing one to form a pairwise constraint. Besides, Abin [20] proposed a method, we call it DQIC (Density-based approach for Querying Informative Constraints), to estimate density and impurity of data points on different adjacency distances and calculate centrality for each data point by applying a density tracking approach on the obtained densities.…”

Section: Active Constraint Learningmentioning

confidence: 99%

See 3 more Smart Citations

Active Block Diagonal Subspace Clustering

Xie

Wang

2021

IEEE Access

Self Cite

View full text Add to dashboard Cite

Subspace clustering aims to find clusters in the low-dimensional subspaces for high-dimensional data. Subspace clustering with Block Diagonal Representation (BDR) maintains the number of connected components of the graph by Laplacian rank constraint, and the learned affinity matrix shows a block diagonal structure, which will achieve a good segmentation for the dataset by spectral clustering. However, the subspaces of real data may overlap and the learned affinity matrix may be imprecise. In this work, we propose an Active learning framework for BDR(ABDR) to acquire and incorporate prior knowledge to improve the subspace clustering performance. An active selection strategy is designed to acquire labels of the informative data points from both the skeleton of clusters and the boundaries of clusters, and then the labeled data are converted into pairwise constraints, which are incorporated into BDR. The optimization of the new objective function is given and the convergence of ABDR is discussed. Experimental results on three images datasets (MNIST, and one UCI dataset(ISOLET) demonstrate the effectiveness of ABDR on complex clustering tasks and show that ABDR is superior to multiple state-of-the-art active clustering and learning techniques.

show abstract

Section: Discussion On Constraint Methodsmentioning

confidence: 99%

Section: A Bdrmentioning

confidence: 99%

Section: B Constrained Subspace Clusteringmentioning

confidence: 99%