7,143 Hits in 5.6 sec

Convergence of the EM Algorithm for Gaussian Mixtures with Unbalanced Mixing Coefficients [article]

Iftekhar Naim, Daniel Gildea (University of Rochester)
2012 arXiv   pre-print
The speed of convergence of the Expectation Maximization (EM) algorithm for Gaussian mixture model fitting is known to be dependent on the amount of overlap among the mixture components.  ...  We propose a deterministic anti-annealing algorithm, that significantly improves the speed of convergence of EM for such mixtures with unbalanced mixing coefficients.  ...  Acknowledgement We would like to thank DanielŠtefankovič, Gaurav Sharma, and Suprakash Datta for many useful comments and feedback. Funded in part by NSF award IIS-0910611.  ... 
arXiv:1206.6427v1 fatcat:ecm7rk2vvfgifmtyox2etocry4

The BYY annealing learning algorithm for Gaussian mixture with automated model selection

Jinwen Ma, Jianfeng Liu
2007 Pattern Recognition  
for the Gaussian mixture.  ...  In order to overcome this difficulty, we propose a simulated annealing learning algorithm to search the global maximum of the harmony function, being expressed as a kind of deterministic annealing EM procedure  ...  Acknowledgments This work was supported by the Natural Science Foundation of China for Projects 60071004, 60471054. The authors thank Prof. Taijun Wang for his simulation supports.  ... 
doi:10.1016/j.patcog.2006.12.028 fatcat:b7q66inpfnccjmyyiwgovshzoi

BayesPy: Variational Bayesian Inference in Python [article]

Jaakko Luttinen
2015 arXiv   pre-print
By removing the tedious task of implementing the variational Bayesian update equations, the user can construct models faster and in a less error-prone way.  ...  It also supports some advanced methods such as stochastic and collapsed variational inference.  ...  An artificial Gaussian mixture dataset is created by drawing 500 samples from two 2-dimensional Gaussian distributions. 200 samples have mean [2, 2] and 300 samples have mean [0, 0]:  ... 
arXiv:1410.0870v3 fatcat:i4pptetpjfgwferh477sh4sv7e

Multi-modal gray-level histogram modeling and decomposition

Jeng-Horng Chang, Kuo-Chin Fan, Yang-Lang Chang
2002 Image and Vision Computing  
In this paper, we present a novel multi-modal histogram thresholding method in which no a priori knowledge about the number of clusters to be extracted is needed.  ...  This prede®ned optimal estimation interval reduces time consumption while other histogram decomposition based methods search all feature space to locate an estimation interval for each candidate cluster  ...  For each cluster C i in Gaussian mixture, we select the interval for initial cluster's mean value estimation by length r i 1=2Vi 2 Vi 2 1 2 1: This length, r i ; possesses the properties of r i q s i and  ... 
doi:10.1016/s0262-8856(01)00095-6 fatcat:gpkhzlge4zazzp7t4lnnyvso2a

Bregman Bubble Clustering: A Robust Framework for Mining Dense Clusters [chapter]

Joydeep Ghosh, Gunjan Gupta
2012 Intelligent Systems Reference Library  
For many problems such as clustering gene-expression datasets where the number of relevant clusters in a dataset are often unknown initially and vary greatly, the more expensive time complexity of the  ...  background results in a set of k dense clusters.  ...  Acknowledgments: This research was supported by NSF grants IIS-0713142 and IIS-1017614. We are also grateful to Srujana Merugu and Arindam Banerjee for some useful discussions.  ... 
doi:10.1007/978-3-642-23166-7_7 fatcat:qigyaoq4arczll3rg32yexrpwu

Bregman bubble clustering

Gunjan Gupta, Joydeep Ghosh
2008 ACM Transactions on Knowledge Discovery from Data  
For many problems such as clustering gene-expression datasets where the number of relevant clusters in a dataset are often unknown initially and vary greatly, the more expensive time complexity of the  ...  background results in a set of k dense clusters.  ...  Acknowledgments: This research was supported by NSF grants IIS-0713142 and IIS-1017614. We are also grateful to Srujana Merugu and Arindam Banerjee for some useful discussions.  ... 
doi:10.1145/1376815.1376817 fatcat:b2ld76mfrrg27hgt54y3ij7xcq

Stability Region Based Expectation Maximization for Model-based Clustering

Chandan Reddy, Hsiao-dong Chiang, Bala Rajaratnam
2006 IEEE International Conference on Data Mining. Proceedings  
To reduce the sensitivity of initial points, a novel algorithm for learning mixture models from multivariate data is introduced in this paper.  ...  Though applied to Gaussian mixtures in this paper, our technique can be easily generalized to any other parametric finite mixture model.  ...  Most of the focus in the literature was on new methods for initialization or new clustering techniques which often do not take advantage of the existing results and completely start the clustering procedure  ... 
doi:10.1109/icdm.2006.152 dblp:conf/icdm/ReddyCR06 fatcat:z3qf4cmhq5gyvaegta3w4k7dhi

Deterministic Initialization of Hidden Markov Models for Human Action Recognition

Zia Moghaddam, Massimo Piccardi
2009 2009 Digital Image Computing: Techniques and Applications  
However, in this paper, we argue that deterministic alternatives are preferable, and propose various methods.  ...  Often, solutions for the selection of the initial parameters are based on random functions.  ...  ∑ = = Σ = M l jl jl jl j N j o G c o b 1 ... 1 , , | µ (3) In (3), µ jl and Σ jl are the mean and covariance of the l-th Gaussian and c jl is its weight in the mixture.  ... 
doi:10.1109/dicta.2009.37 dblp:conf/dicta/MoghaddamP09 fatcat:3cpdsnf6undeba5vwuw32kxcti

Color Image Segmentation Using Adaptive Spatial Gaussian Mixture Model

M.Sujaritha, S. Annadurai
2010 Zenodo  
An adaptive spatial Gaussian mixture model is proposed for clustering based color image segmentation.  ...  The experimental results with synthetic and Berkeley dataset demonstrate that the proposed method is effective in improving the segmentation and it can be employed in different practical image content  ...  Boundarybased methods search for the most dissimilar pixels which represent discontinuities in the image, while region based methods on the contrary search for the most similar areas.  ... 
doi:10.5281/zenodo.1334283 fatcat:w6ddhx72z5dv7fsc3yonmwtlku

Reset-Free Guided Policy Search: Efficient Deep Reinforcement Learning with Stochastic Initial States [article]

William Montgomery, Anurag Ajay, Chelsea Finn, Pieter Abbeel, Sergey Levine
2016 arXiv   pre-print
We compare our method to existing policy search techniques in simulation, showing that it can train high-dimensional neural network policies with the same sample efficiency as prior GPS methods, and present  ...  In contrast to prior GPS methods, which require a consistent set of initial states to which the system must be reset after each episode, our approach can handle randomized initial states, allowing it to  ...  to each of K clusters, producing datasets {D k } 4: while cluster assignments not converged do 5: Fit each linear-Gaussian dynamics p k (x t+1 |x t , u t ) using samples in each D k 6: Fit each linearized  ... 
arXiv:1610.01112v2 fatcat:5t75jq6ctrb3fkg2rg3wfymsoy

Automatic Quantification of Fluorescence from Clustered Targets in Microscope Images [chapter]

Harri Pölönen, Jussi Tohka, Ulla Ruotsalainen
2009 Lecture Notes in Computer Science  
A cluster of fluorescent targets appears as overlapping spots in microscope images. By quantifying the spot intensities and locations, the properties of the fluorescent targets can be determined.  ...  Commonly this is done by reducing noise with a low-pass filter and separating the spots by fitting a Gaussian mixture model with a local optimization algorithm.  ...  JT received additional support from University Alliance Finland Research Cluster of Excellence STATCORE. HP received additional support from Jenny and Antti Wihuri Foundation.  ... 
doi:10.1007/978-3-642-02230-2_68 fatcat:zfj5ehzg6bfndfit2fxaguqrkq

Newtonian clustering: An approach based on molecular dynamics and global optimization

K. Blekas, I.E. Lagaris
2007 Pattern Recognition  
Further refinement is achieved by applying the EM-algorithm to a Gaussian mixture model whose construction and initialization is based on the acquired information.  ...  Most clustering methods assume that this number is a priori known, and then try to place K clusters in the space defined by the data.  ...  Acknowledgments We wish to thank the anonymous referee for the valuable comments that significantly improved the clarity of the article, and our colleagues, Professors Galatsanos and Likas for illuminating  ... 
doi:10.1016/j.patcog.2006.07.012 fatcat:vwgcduwv4zb4vhdnvj5ga2jn74

Color image segmentation using adaptive mean shift and statistical model-based methods

Jong Hyun Park, Guee Sang Lee, Soon Young Park
2009 Computers and Mathematics with Applications  
The number of mixture components is determined automatically by adaptive mean shift, in which local clusters are estimated by repeatedly searching for higher density points in feature vector space.  ...  algorithm for color images based on Gaussian mixture models (GMMs).  ...  The adaptive mean shift has been used to determine the number of components in a Gaussian mixture model and to detect the modes of each mixture component.  ... 
doi:10.1016/j.camwa.2008.10.053 fatcat:ifshfnvjlzcj7lephnt4dsu5sq

The Role of Hubness in Clustering High-Dimensional Data

Nenad Tomasev, Milos Radovanovic, Dunja Mladenic, Mirjana Ivanovic
2014 IEEE Transactions on Knowledge and Data Engineering  
The proposed methods are tailored mostly for detecting approximately hyperspherical clusters and need to be extended in order to properly handle clusters of arbitrary shapes.  ...  High-dimensional data arise naturally in many domains, and have regularly presented a great challenge for traditional data-mining techniques, both in terms of effectiveness and efficiency.  ...  This work was supported by the Slovenian Research Agency, the IST Programme of the EC under PASCAL2 (IST-NoE-216886), and the Serbian Ministry of Education, Science and Technological Development project  ... 
doi:10.1109/tkde.2013.25 fatcat:qmos3ipvbjfbld47talxbllpzi

Examining the effect of initialization strategies on the performance of Gaussian mixture modeling

Emilie Shireman, Douglas Steinley, Michael J. Brusco
2015 Behavior Research Methods  
Like many approaches for identifying subpopulations, finite mixture modeling can suffer from locally optimal solutions, and the final parameter estimates are dependent on the initial starting values of  ...  Generally, the parameters of these Gaussian mixtures cannot be estimated in closed form, so estimates are typically obtained via an iterative process.  ...  In the common case of mixtures of multivariate normal distributions (e.g., Gaussian mixture modeling), each f k is the Gaussian density, and ϑ k contain the p × 1 mean vector μ k and the p × p covariance  ... 
doi:10.3758/s13428-015-0697-6 pmid:26721666 pmcid:PMC4930421 fatcat:tyyjuivga5hrbnczx4itbqqg5y
« Previous Showing results 1 — 15 out of 7,143 results