Unified clustering comparison framework for overlapping and hierarchical structures.
problem Critical biases in existing clustering comparison measures.
method Element-centric framework comparing relationships induced by cluster structure.
result Framework does not suffer from biases and provides unique insights.
Paper tackles clustering with ordinal comparisons, achieving near-optimal results.
problem Clustering with ordinal comparisons when similarity measures are not available.
method Two-step procedure: estimate similarity matrix from comparisons, then apply SDP clustering.
result Near-optimal recovery of planted clustering using near-optimal number of comparisons.
Develops comparison-based hierarchical clustering algorithms without object representations.
problem Hierarchical clustering without object representations or pairwise similarities.
method Comparison-based hierarchical clustering algorithms (single, complete, and average linkage variants).
result Statistical guarantees and empirical performance on various datasets.
Proposes a revenue function to evaluate dendrograms from comparisons.
problem Evaluate dendrograms from comparisons without ground-truth.
method Introduces a new revenue function related to Dasgupta's cost.
result Revenue function allows meaningful evaluation of dendrograms.
Paper tackles noisy comparison oracle for robust clustering algorithms.
problem Finding robust clustering algorithms under noisy comparison oracle.
method Develops algorithms for k-center clustering and agglomerative hierarchical clustering using noisy comparison oracle.
result Proves robust algorithms achieve good approximation guarantees with high probability.
Benchmarked over 70 graph clustering algorithms.
problem Lack of comprehensive performance comparison for graph clustering algorithms.
method Evaluated 70+ graph clustering programs for runtime and quality on weighted and unweighted graphs, analyzed ground truth characteristics.
result Supply a start point for engineers and viewpoint for researchers.
Given a set of pairwise comparisons, the classical ranking problem computes a single ranking that best represents the preferences of all users. In this paper, we study the problem of inferring individual preferences, arising in the context of making personalized recommendations. In particular, we assume that there are …
A new clustering algorithm fuses heat diffusion and turning angle for robustness.
problem Cluster similar elements in various fields.
method Combines heat diffusion and maximal turning angle for robust fission clustering.
result The SARFC algorithm outperforms other methods in clustering performance.
An algorithm learns a kernel matrix from relative-distance constraints for semi-supervised clustering.
problem Learning metrics from relative-distance constraints to capture finer structures.
method Log determinant divergence for kernel matrix learning with relative-distance constraints.
result Kernels learned from relative-distance constraints yield better clusterings than existing methods.
FastAMI efficiently approximates AMI and SMI for large datasets.
problem Computational difficulty in comparing clusterings with an adjustment for chance.
method Monte Carlo-based approach to approximate AMI and SMI.
result FastAMI provides accurate results for large datasets.
Study compares clustering methods for student poverty levels in unsupervised surveys.
problem Identifying impoverished students in unsupervised survey data.
method Multiple clustering techniques (k-means, k-modes, hierarchical clustering) applied to student survey data.
result Fuzzy logic used for data cleaning and organizing, identifying most viable clustering method for survey data.
Active learning optimizes correlation clustering by querying the most informative pairwise comparisons.
problem Efficiently clustering data with limited pairwise similarity information.
method Developed principled active learning approach using information-theoretic acquisition functions.
result Significantly outperforms existing baselines in clustering accuracy and query efficiency.
This paper compares two clustering evaluation metrics, revealing their differences and properties.
problem Understanding the differences between misclassification error distance and adjusted Rand index.
method Population origins, data analysis examples, detailed case studies, and simulation study.
result Reveals previous misconceptions about the two metrics and their distributions.
ECG improves graph clustering and resolves resolution limit issues.
problem Graph clustering resolution limit issue.
method ECG uses consensus clustering to improve graph clustering.
result ECG alleviates the resolution limit issue and improves partition stability.
New random models improve clustering similarity assessment.
problem Improper random models affect clustering similarity assessments.
method Derived corrected Rand index and Mutual Information measures for varying cluster sizes.
result Random model choice drastically impacts clustering similarity rankings.
C-FAR automates clustering assessment for neural tracking.
problem Manual assessment of clusters by humans is slow and impractical for large datasets.
method C-FAR uses automated feedback queries to select optimal clustering from multiple algorithms.
result C-FAR produces near-perfect clustering on simulated neural data.
We formulate weighted graph clustering as a prediction problem: given a subset of edge weights we analyze the ability of graph clustering to predict the remaining edge weights. This formulation enables practical and theoretical comparison of different approaches to graph clustering as well as comparison of graph cluste…
This article reviews and compares clustering algorithms for large datasets.
problem Detecting patterns in large unlabeled data.
method Examines and compares various clustering algorithms.
result Shows strengths and weaknesses of clustering techniques based on dataset size.
Paper tackles learning mixture of RUMs from partial data.
problem Learning a mixture of Random Utility Models (RUMs) from pairwise comparisons.
method PCA-based spectral clustering to reduce mixture to single component.
result Algorithm correctly clusters data from a mixture of RUMs with high probability.
Paper compares clusterability measures for data mining.
problem Selecting appropriate clusterability measures for data.
method Extensive comparison of clusterability measures.
result Guidelines for selecting suitable measures for clustering.
New method identifies common organizational principles in networks.
problem Challenging to cluster networks of different size and density.
method Introduces a new network comparison methodology.
result Identifies common organizational principles in networks.
We examine methods for clustering in high dimensions. In the first part of the paper, we perform an experimental comparison between three batch clustering algorithms: the Expectation-Maximization (EM) algorithm, a winner take all version of the EM algorithm reminiscent of the K-means algorithm, and model-based hierarch…
A new metric optimizes clustering and compares results from various methods.
problem Determining the right number of clusters and comparing different clustering methods.
method Proposes a novel metric to find the optimal number of clusters and compare different clustering techniques.
result Optimizes clustering and cross-comparison of results from different methods.
A deep generative model with a relational model tackles noisy pairwise comparisons for semi-supervised clustering.
problem Noisy pairwise comparisons on a small subset of data for clustering.
method Deep generative model (DGM) and statistical relational model, sharing latent variables, Bayesian variant, stochastic variational inference.
result Outperforms previous crowdsourced clustering methods on synthetic and real-world datasets.
Adjusted for chance measures are widely used to compare partitions/clusterings of the same data set. In particular, the Adjusted Rand Index (ARI) based on pair-counting, and the Adjusted Mutual Information (AMI) based on Shannon information theory are very popular in the clustering community. Nonetheless it is an open …
Intuitive clustering algorithm balances cluster size and cohesion.
problem Cluster definition and selection in data analysis.
method Nearest neighbours equilibrium condition for clustering.
result High-quality clustering solutions compared to benchmarks.
The paper analyzes indices based on counting object pairs for assessing partition agreement in unsupervised learning.
problem The difficulty in interpreting overall indices like Rand and adjusted Rand indices.
method Analysis of three families of indices based on counting object pairs, decomposing overall indices into cluster-level indices.
result Overall indices based on pair-counting approach are sensitive to cluster size imbalance and provide limited information on smaller clusters.
We explore the performance of several automatic bandwidth selectors, originally designed for density gradient estimation, as data-based procedures for nonparametric, modal clustering. The key tool to obtain a clustering from density gradient estimators is the mean shift algorithm, which allows to obtain a partition not…
Transforms data into separable subspaces for clustering.
problem Data is not always separable into subspaces.
method Embeds subspace clustering techniques into transform learning.
result Improves upon state-of-the-art clustering techniques.
Proposes SAG-DBSCAN for clustering with self-adaptation.
problem Clustering analysis in data mining.
method Uses grey relational matrix and DBSCAN for clustering.
result Demonstrates superior performance compared to other methods.
New clustering method using point-set kernel measures similarity.
problem Measuring similarity between objects for clustering.
method Point-set kernel for similarity computation; clustering procedure uses this measure.
result Proposed method is more effective and faster than existing algorithms.
New visual quality index for fuzzy clustering.
problem No accurate quality index for fuzzy clustering across datasets.
method Proposes a new visual quality index and graph-based solution.
result Validated through extensive experiments on various datasets.
A new co-clustering model for high-dimensional data reduces parameter complexity.
problem High-dimensional data challenges traditional co-clustering methods.
method Parameter-wise co-clustering model with SEM and Gibbs sampler for estimation.
result The model maintains parsimony while offering more flexibility.
EAP clusters evolving data, promoting temporal smoothness and automatic cluster tracking.
problem Clustering time-evolving data with temporal smoothness and automatic cluster identification.
method Evolutionary Affinity Propagation (EAP) on a factor graph exchanging messages between adjacent data snapshots.
result EAP clusters data with temporal smoothness and automatically tracks clusters, outperforming existing methods.
This paper analyzes various graph clustering methods and their applications.
problem Dividing graphs into homogeneous groups for diverse applications.
method Traditional and deep learning-based clustering methods are compared.
result Deep learning techniques improve clustering accuracy.
Much of the data being created on the web contains interactions between users and items. Stochastic blockmodels, and other methods for community detection and clustering of bipartite graphs, can infer latent user communities and latent item clusters from this interaction data. These methods, however, typically ignore t…
This paper introduces a novel model-based clustering approach for clustering time series which present changes in regime. It consists of a mixture of polynomial regressions governed by hidden Markov chains. The underlying hidden process for each cluster activates successively several polynomial regimes during time. The…
Paper uses GMM with DVAE to detect star clusters in noisy images.
problem Detecting stellar clusters in astronomical images.
method Unsupervised approach using Deep Variational Autoencoder (DVAE) combined with Gaussian Mixture Model (GMM).
result Method outperforms state-of-the-art in recognizing star clusters, even in noisy images.
Proposes a new clustering algorithm using random forest.
problem Density-based clustering with optimal level determination.
method Best-scored random forest algorithm.
result Guaranteed consistency and fast convergence rates.
This paper tackles deep clustering evaluation challenges in high-dimensional data.
problem Evaluation of deep clustering methods is problematic due to the curse of dimensionality and variations in embedding spaces.
method Develops a theoretical framework to highlight the ineffectiveness of internal validation measures and proposes a systematic approach to applying clustering validity indices in deep learning.
result The proposed framework reduces misguidance from improper use of clustering validity indices in deep learning.
Efficient clustering for large datasets using a sampling-based approach.
problem Clustering high-dimensional data with a large number of clusters efficiently.
method A simple and efficient clustering method that evaluates distances of data points with a subset of cluster centers.
result Optimal solutions of the approximation are the same as in the exact solution, but more efficient at extracting clusters.
A method for clustering using transfer learning from similar labeled data.
problem Clustering with datasets having different features and labeled data.
method Constructing meta-features to describe structural characteristics of data and transferring them between source and target domains.
result The method is efficient and works under arbitrary feature descriptions of source and target domains with smaller complexity.
Proposes a new hierarchical clustering method combining DP and DBSCAN strengths.
problem Combining strengths of DP and DBSCAN for arbitrary shape clusters.
method DC-HDP: Combines Density Peak and Density-Connectivity approaches.
result Produces best clustering results on 14 datasets.
Two methods using low-discrepancy points improve data compression for neural networks.
problem Efficiently compress large datasets for neural network training.
method Two methods based on low-discrepancy points: digital nets with averaging and clustering.
result Second method outperforms supercompress in compression error and neural network accuracy.
This paper proposes a variant of the method of Guédon and Verhynin for estimating the cluster matrix in the Mixture of Gaussians framework via Semi-Definite Programming. A clustering oriented embedding is deduced from this estimate. The procedure is suitable for very high dimensional data because it is based on pairwis…
A new sliced IGW distance for Gromov-Wasserstein alignment.
problem Scalability issues in Gromov-Wasserstein alignment for high-dimensional problems.
method Proposed a sliced IGW distance with rotational invariance.
result Natural rotational invariance of the sliced IGW distance.
Unsupervised clustering of curves according to their shapes is an important problem with broad scientific applications. The existing model-based clustering techniques either rely on simple probability models (e.g., Gaussian) that are not generally valid for shape analysis or assume the number of clusters. We develop an…
New clustering methods for binary data using combinatorial optimization.
problem Clustering binary data efficiently and effectively.
method Five new combinatorial optimization heuristics (SA, TA, TS, GA, ACO) applied to binary data.
result Simulated annealing performs exceptionally well compared to classical methods.