Enhanced Ensemble Clustering via Fast Propagation of Cluster-Wise Similarities

Abstract

Ensemble clustering has been a popular research topic in data mining and machine learning. Despite its significant progress in recent years, there are still two challenging issues in the current ensemble clustering research. First, most of the existing algorithms tend to investigate the ensemble information at the object-level, yet often lack the ability to explore the rich information at higher levels of granularity. Second, they mostly focus on the direct connections (e.g., direct intersection or pairwise co-occurrence) in the multiple base clusterings, but generally neglect the multi-scale indirect relationship hidden in them. To address these two issues, this paper presents a novel ensemble clustering approach based on fast propagation of cluster-wise similarities via random walks. We first construct a cluster similarity graph with the base clusters treated as graph nodes and the cluster-wise Jaccard coefficient exploited to compute the initial edge weights. Upon the constructed graph, a transition probability matrix is defined, based on which the random walk process is conducted to propagate the graph structural information. Specifically, by investigating the propagating trajectories starting from different nodes, a new cluster-wise similarity matrix can be derived by considering the trajectory relationship. Then, the newly obtained cluster-wise similarity matrix is mapped from the cluster-level to the object-level to achieve an enhanced co-association (ECA) matrix, which is able to simultaneously capture the object-wise co-occurrence relationship as well as the multi-scale cluster-wise relationship in ensembles. Finally, two novel consensus functions are proposed to obtain the consensus clustering result. Extensive experiments on a variety of real-world datasets have demonstrated the effectiveness and efficiency of our approach.

Ensemble cluster pruning via convex‐concave programming

“…Their results show an improvement from the perspective of quality and efficiency . Another ensemble clustering method is adapted from the fast propagation algorithm of cluster‐wise similarities by random walks . In the study of Huang et al, a new cluster‐wise similarity measure, an enhanced coassociation matrix and two new consensus functions are proposed to improve the performance of the ensemble clustering approach.…”

“…Another ensemble clustering method is adapted from the fast propagation algorithm of cluster‐wise similarities by random walks . In the study of Huang et al, a new cluster‐wise similarity measure, an enhanced coassociation matrix and two new consensus functions are proposed to improve the performance of the ensemble clustering approach. Moreover, in the literature, an automatic clustering ensemble selection method is introduced to prune the random forest classifier according to clustering and selection steps .…”

Ant Colony Clustering for ROI Identification in Functional Magnetic Resonance Imaging

“…be the HRF convolution matrix of size T × T (4) for all ℓ ∈ L do (5) Initialize the parameters α, k 1 , k 2 (6) Compute the UMAP grid using as inputs the neuronal stimulus estimated from fMRI signals. Assign the ants to this UMAP grid (7) Estimate the neuronal stimulus by deconvolution using H [26,27] (8) for j � 1 to K L do (9) for t � 1 to t max do (10) for all ant ∈ A ℓ do (11) if ant unladel and grid occupied by e ℓ j i , then (12) Compute f(e ℓ j i ) and P p according to equations (2) and 3, respectively (13) Generate a random number p ∈ [0, 1] 14if p ≤ P p (e ℓ j i ), then (15) Pick up the neuronal signal e ℓ j i (16) end if (17) else (18) if ant carrying neuronal signal e ℓ j i and grid empty (19) Compute f(e ℓ j i ) and P d according to equations (2) and 4, respectively (20) Generate a random number p…”

“…Among existing clustering-based parcellation techniques, there are a number of works applying k-means variants [8,12,13], mixture models [11], hierarchical clustering [14,15], and spectral clustering [10]. Despite the above approaches have been extensively used in solving the task of finding ROIs in fMRI, more modern ideas from the clustering literature can be incorporated to enhance performance, robustness, and efficiency [16][17][18][19].…”

Spectral Clustering by Subspace Randomization and Graph Fusion for High-Dimensional Data

“…Data clustering is a fundamental yet still very challenging problem in data mining and knowledge discovery [13]. A large number of clustering techniques have been developed in the past few decades [2][3][4][5][6][8][9][10][11][12][14][15][16][17][18][21][22][23][24], out of which the spectral clustering has been a very important category with its effectiveness and robustness in dealing with complex data [3,6,14,18,22]. In this paper, we focus on the spectral clustering technique, especially for high-dimensional scenarios.…”