Parsimonious Gaussian mixture models

McNicholas, Paul D.; Murphy, Brendan

doi:10.1007/s11222-008-9056-0

Cited by 282 publications

(286 citation statements)

References 26 publications

Supporting

Mentioning

284

Contrasting

Order By: Relevance

“…The clustering is achieved by fitting a mixture model to the latent Gaussian data. The model is closely related to the mixture of factor analysers model [14,15] for continuous data; in the case of the mixture of item response models however, only a discrete version of the data are observed. Bridge sampling was employed for model selection.…”

Section: Discussionmentioning

confidence: 99%

Clustering Ordinal Data via Latent Variable Models

McParland

Gormley

2013

Algorithms From and for Nature and Life

View full text Add to dashboard Cite

Section: Discussionmentioning

confidence: 99%

Clustering Ordinal Data via Latent Variable Models

McParland

Gormley

2013

Algorithms From and for Nature and Life

View full text Add to dashboard Cite

“…Mixture Models (PGMM) method was used for cluster analysis of differentially expressed genes 53 [181][182][183]. In addition, k-means, k-medoids, and MCLUST clustering methods were used to subdivide larger PGMM clusters in an effort to highlight smaller gene networks [184][185][186][187][188].…”

Section: Microarray Cluster Analysismentioning

confidence: 99%

Elucidating the roles of stearoyl-CoA desaturase 1 in adipocyte fatty acid metabolism and cellular function

Ralston

2015

Appl. Physiol. Nutr. Metab.

View full text Add to dashboard Cite

Worldwide obesity rates have risen to epidemic proportions, with over 600 million obese individuals across the globe. These individuals are prone to obesity-related health complications including type 2 diabetes, hypertension, cancer and cardiovascular disease. Obesity also coincides with the expansion of adipose tissue (AT), which has an important role storing excess calories in the form of triacylglycerol (TAG) within adipocyte lipid droplets. The predominant fatty acids (FAs) comprising adipocyte TAGs are monounsaturated FAs (MUFAs), which are produced by stearoyl-CoA desaturase 1 (SCD1). Specifically, SCD1 converts saturated FAs palmitate (PA) and stearate (SA) into palmitoleate (PMA) and oleate (OA), respectively.Interestingly, whole-body SCD1-deficiency is associated with reduced lipogenesis and adiposity.This places SCD1 as a potential target for obesity therapies; however, our understanding of the mechanisms linking SCD1 with changes in adipocyte function remains unclear. This thesis provides important new insights into how SCD1 impacts adipocyte FA metabolism and cellular function. Using a specific SCD1 inhibitor, changes in lipid metabolism, global gene expression, inflammation, cellular stress and basal insulin signaling were assessed in 3T3-L1 adipocytes.Results demonstrated that SCD1 inhibition caused a reduction in TAGs and phospholipids, which coincided with the down-regulation of genes associated with the biosynthesis of these lipid fractions. Cellular diacylglyerols were increased with SCD1 inhibition and insulin signaling was partially impaired. In contrast, markers of cellular stress were unaltered. Furthermore, the FA composition of each lipid fraction was dramatically modified, with SCD1-inhibited adipocytes specifically up-regulating the elongation of PA to SA. Stable isotope tracer experiments revealed that this elongation was occurring via elongase 6. Additionally, reduced SCD1 activity in adipocytes exacerbated the effects of exogenous SA on markers or inflammation. Taken together, SCD1 activity has many indirect influences on adipocyte metabolism in addition to its role in FA desaturation. Importantly, SCD1 facilitates the storage of FAs in TAGs and the ability of adipocytes to handle exogenous FAs. SCD1 also prevents saturated FA and DAG accumulation, and preserves insulin signaling in adipocytes. Ultimately this thesis highlights the importance of SCD1 in the maintenance of adipocyte cellular function, and emphasizes the wide-ranging impact of SCD1 on adipocyte FA metabolism.iv

show abstract

“…A general framework for the MFA model was also proposed by McNicholas & Murphy [54] which includes the previous works of Ghahramani and Hinton and of McLachlan et al [52]. By considering the previous framework, defined by Equations (4) and (6), McNicholas & Murphy [55] proposed a family of models known as the expanded parsimonious Gaussian mixture model (EPGMM) family.…”

Section: Mixture Of Parsimonious Gaussian Mixture Models (Pgmm)mentioning

confidence: 99%

“…They decline 12 EPGMM models by either constraining the terms of the covariance matrix to be equal or not, considering an isotropic variance for the noise term, or re-parametrizing the factor analysis covariance structure. According to this family of 12 models, the previous approaches developed by [3,33,52,54,71] then become sub-models of the EPGMM approach. For example, by constraining only the noise variance to be isotropic on each class (Ψ k = σ 2 k I p ), which by the way corresponds to the UUC and UUUC models, it produces the famous mixture of probabilistic PCA (Mixt-PPCA) of Tipping & Bishop [71].…”

Section: Mixture Of Parsimonious Gaussian Mixture Models (Pgmm)mentioning

confidence: 99%

“…The UUUU model is equivalent to the MFA model proposed by McLachlan et al in [52]. Finally, by parameterizing the covariance structure as Ψ k = ω k ∆ k , where ∆ k is a diagonal matrix and |∆ k | = 1, McNicholas & Murphy proposed 4 additional models to their previous work [54]. In the case of the EPGMM models, McNicholas & Murphy [55] proposed in the clustering context to make use of the AECM algorithm for the model inference.…”

Section: Mixture Of Parsimonious Gaussian Mixture Models (Pgmm)mentioning

confidence: 99%

See 1 more Smart Citation

Probabilistic model‐based discriminant analysis and clustering methods in chemometrics

Bouveyron¹

2013

Journal of Chemometrics

View full text Add to dashboard Cite

In Chemometrics, the supervised and unsupervised classification of high‐dimensional data has become a recurrent problem. Model‐based techniques for discriminant analysis and clustering are popular tools which are renowned for their probabilistic foundations and their flexibility. However, classical model‐based techniques show a disappointing behavior in high‐dimensional spaces which up to now have been limited in their use within Chemometrics. The recent developments in model‐based classification overcame these drawbacks and enabled the efficient classification of high‐dimensional data. This work presents a comprehensive review of these recent approaches, including regularization‐based techniques, parsimonious modeling, subspace classification methods and classification methods based on variable selection.

show abstract

Parsimonious Gaussian mixture models

Cited by 282 publications

References 26 publications

Clustering Ordinal Data via Latent Variable Models

Clustering Ordinal Data via Latent Variable Models

Elucidating the roles of stearoyl-CoA desaturase 1 in adipocyte fatty acid metabolism and cellular function

Probabilistic model‐based discriminant analysis and clustering methods in chemometrics

Contact Info

Product

Resources

About