Article
Operations Research & Management Science
Tommaso Rigon
Summary: This paper introduces a novel enriched Dirichlet mixture model for clustering functional data, incorporating functional constraints and bounding the model complexity. The prior process is characterized through a urn scheme, enhancing the interpretability of clustering. Variational Bayes approximation is employed for tractable posterior inference, overcoming computational bottlenecks.
APPLIED STOCHASTIC MODELS IN BUSINESS AND INDUSTRY
(2023)
Article
Engineering, Environmental
Asael Fabian Martinez, Somnath Chaudhuri, Carlos Diaz-Avalos, Pablo Juan, Jorge Mateu, Ramses H. Mena
Summary: An unsupervised classification method is proposed for point events occurring on a geometric network. It utilizes the flexibility and practicality of random partition models to discover clustering structures of observations from a specific phenomenon on a given set of edges. By incorporating spatial effects through a random partition distribution induced by a Dirichlet process, the method offers an appealing clustering approach. A Gibbs sampler algorithm is proposed and evaluated with sensitivity analysis. The analysis of crime and violence patterns in Mexico City serves as the motivation and illustration for this proposal.
STOCHASTIC ENVIRONMENTAL RESEARCH AND RISK ASSESSMENT
(2023)
Article
Mathematics, Interdisciplinary Applications
Daniel Ayala, Leonardo Jofre, Luis Gutierrez, Ramses H. Mena
Summary: This paper introduces an explicit representation of phase-type distributions as an infinite mixture of Erlang distributions, revealing a novel and useful connection between a class of Bayesian nonparametric mixture models and phase-type distributions. The paper explores estimation techniques for phase-type distributions and closed-form expressions for functionals related to Dirichlet process mixture models, and demonstrates the power of this connection through a posterior inference algorithm.
Article
Statistics & Probability
George Karabatsos
Summary: The article introduces a new fast-search algorithm for Bayesian nonparametric (BNP) infinite-mixture models, which can handle a wide range of BNP priors and is efficient in processing large datasets.
JOURNAL OF COMPUTATIONAL AND GRAPHICAL STATISTICS
(2021)
Article
Computer Science, Artificial Intelligence
Andreas Wichert
Summary: A new concept of a quantum-like mixture Gaussian model is introduced to describe the mixture distribution by assuming a point is generated by each Gaussian simultaneously. This model improves classification accuracy in machine learning by indicating that uncertain points should not be assigned to any class, increasing the accuracy on the iris data set from 96.67% to 99.24%.
Article
Ecology
Denis Valle, Yusuf Jameel, Brenda Betancourt, Ermias T. Azeria, Nina Attias, Joshua Cullen
Summary: This article highlights the importance and advantages of using Bayesian clustering methods in ecological and environmental sciences, and proposes the use of sparsity-inducing priors to determine the number of groups. Through application examples using simulated and real data, it demonstrates that this approach can successfully recover the true number of groups.
ECOLOGICAL APPLICATIONS
(2022)
Article
Computer Science, Information Systems
Guanli Yue, Ansheng Deng, Yanpeng Qu, Hui Cui, Xueying Wang
Summary: This paper proposes a novel spectral clustering algorithm that addresses the limitations of existing algorithms by incorporating density stratification and density ratio to handle complex multi-density data.
INFORMATION SCIENCES
(2023)
Article
Engineering, Industrial
Abdallah Chehade, Mayuresh Savargaonkar, Vasiliy Krivtsov
Summary: This work addresses the challenge of warranty data maturation in forecasting warranty claims for complex products by proposing the Conditional Gaussian Mixture Model (CGMM), which utilizes historical warranty data to develop a robust prior joint distribution of warranty trends and estimates the posterior distribution of warranty claims at future maturation levels. The CGMM is validated on a large automotive warranty claims dataset and effectively identifies non-parametric temporal warranty trends and clusters products into latent groups.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2022)
Article
Biology
Michael J. Daniels, Minji Lee, Wei Feng
Summary: In longitudinal studies, multiple attempts to collect a measurement after baseline are common. Recording whether these attempts are successful is important for assessing missing data assumptions. Previous models were limited in their ability to perform sensitivity analysis, but we propose a new approach that uses Bayesian nonparametrics and introduces a novel method for identification and sensitivity analysis. We applied this approach to a clinical trial dataset and conducted simulations to evaluate its properties.
Article
Computer Science, Artificial Intelligence
Aristeidis Panos, Petros Dellaportas, Michalis K. Titsias
Summary: This study introduces a Gaussian process latent factor model for multi-label classification, which can capture correlations among class labels. To address computational challenges, several techniques are introduced, including variational sparse Gaussian process and stochastic optimization. The results demonstrate the practicality of this method in large-scale multi-label learning problems.
Article
Environmental Sciences
Luyang Wang, Jinhui Lan
Summary: In this study, a segmentation method based on a roadside LiDAR sensor is proposed to efficiently and accurately segment vehicles and pedestrians on urban roads. By constructing a polar grid and using an adaptive polar-grid Gaussian-mixture model, the accuracy of the segmentation is improved. Additionally, a density-adaptive DBSCAN target-clustering algorithm is used to solve the problem of low clustering accuracy caused by uneven point cloud density.
Article
Social Sciences, Mathematical Methods
Leonardo Egidi, Francesco Pauli, Nicola Torelli, Susanna Zaccarin
Summary: We propose a Bayesian model-based clustering technique that incorporates network relations and geographical positioning of territorial units. The aim is to design administrative structures in an Italian region based on commuting flows between municipalities. The social network model explains commuting flows using distances between municipalities in a 3-dimensional space modeled through a Gaussian mixture.
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES A-STATISTICS IN SOCIETY
(2023)
Article
Computer Science, Theory & Methods
Marco Stefanucci, Antonio Canale
Summary: This paper introduces a novel family of multiscale stick-breaking mixture models that combine the advantages of single-scale nonparametric mixtures and Polya trees, suitable for estimating densities with varying degrees of smoothness and local features.
STATISTICS AND COMPUTING
(2021)
Article
Engineering, Civil
Xiaoxu Chen, Chengyuan Zhang, Zhanhong Cheng, Yuang Hou, Lijun Sun
Summary: Car-following models are essential for microscopic traffic simulation. This study presents a data-driven model based on a Bayesian Gaussian mixture model for probabilistic forecasting of human car-following behaviors. The model captures the temporal dynamics of human car-following behaviors and provides accurate predictions with quantified uncertainty. The results suggest that this model is promising for modeling and forecasting car-following behaviors.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Automation & Control Systems
Adriana Laura Lopez-Lobato, Martha Lorena Avendano-Garrido
Summary: A method for estimating the parameters of a Gaussian mixture model's density function by minimizing the Gini index between an empirical data distribution and the model was proposed in this paper. The method was validated through simulated and real data examples, demonstrating its effectiveness and properties.
INTERNATIONAL JOURNAL OF APPLIED MATHEMATICS AND COMPUTER SCIENCE
(2021)
Article
Chemistry, Analytical
Reham Badawy, Yordan P. Raykov, Luc J. W. Evers, Bastiaan R. Bloem, Marjan J. Faber, Andong Zhan, Kasper Claes, Max A. Little
Article
Multidisciplinary Sciences
Yordan P. Raykov, Alexis Boukouvalas, Fand Baig, Max A. Little
Article
Chemistry, Analytical
Yazan Qarout, Yordan P. Raykov, Max A. Little
Article
Genetics & Heredity
Monika Krzak, Yordan Raykov, Alexis Boukouvalas, Luisa Cutillo, Claudia Angelini
FRONTIERS IN GENETICS
(2019)
Article
Health Care Sciences & Services
Luc J. W. Evers, Yordan P. Raykov, Jesse H. Krijthe, Ana Ligia Silva de Lima, Reham Badawy, Kasper Claes, Tom M. Heskes, Max A. Little, Marjan J. Meinders, Bastiaan R. Bloem
JOURNAL OF MEDICAL INTERNET RESEARCH
(2020)
Article
Acoustics
Amir Hossein Poorjam, Mathew Shaji Kavalekalam, Liming Shi, Jordan P. Raykov, Jesper Rindom Jensen, Max A. Little, Mads Graesboll Christensen
Summary: This study investigates the impact of various acoustic degradations on the performance of voice-based Parkinson's disease detection systems, and proposes two methods for automatically controlling the quality of recordings to improve PD detection accuracy. Experimental results demonstrate the effectiveness of quality control approaches in selecting appropriate enhancement methods, leading to improved PD detection accuracy.
SPEECH COMMUNICATION
(2021)
Article
Computer Science, Information Systems
Yordan P. Raykov, Luc J. W. Evers, Reham Badawy, Bastiaan R. Bloem, Tom M. Heskes, Marjan J. Meinders, Kasper Claes, Max A. Little
Summary: In this study, a principled modeling approach for free-living gait analysis was developed to support health predictions and clinical diagnosis. Using a dataset of PD patients and controls, the framework's effectiveness in detecting gait and predicting medication-induced fluctuations in PD patients was demonstrated. The approach was shown to be robust to varying sensor locations.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2021)
Article
Engineering, Electrical & Electronic
Yordan P. Raykov, David Saad
Summary: This paper introduces the fundamental concepts of some commonly used machine learning methods, excluding deep-learning machines and neural networks. It discusses their advantages, limitations, and potential applications in various fields of photonics. The main methods covered include parametric and nonparametric regression and classification techniques, kernel-based methods, support vector machines, decision trees, probabilistic models, Bayesian graphs, mixture models, Gaussian processes, message passing methods, and visual informatics.
IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS
(2022)
Proceedings Paper
Acoustics
Amir Hossein Poorjam, Yordan P. Raykov, Reham Badawy, Jesper Rindom Jensen, Mads Graesboll Christensen, Max A. Little
2019 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP)
(2019)