Independent component analysis and clustering for pollution data

Chattopadhyay, Asis Kumar; Mondal, Saptarshi; Biswas, Atanu

doi:10.1007/s10651-014-0287-2

Cited by 7 publications

(4 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the fastICA, each component is treated with equal importance, whereas in PCA, some attributes are measured with high importance [19]. The considered elements' statistical autonomy is maximized for finding the independent components of fastICA [18]. Independence in fastICA is achieved by either maximizing the non-gaussianity or Minimizing Mutual Information (MMI) [91].…”

Section: B Fast Independent Component Analysismentioning

confidence: 99%

See 1 more Smart Citation

bSSA: Binary Salp Swarm Algorithm With Hybrid Data Transformation for Feature Selection

et al. 2021

View full text Add to dashboard Cite

Section: B Fast Independent Component Analysismentioning

confidence: 99%

“…In continuation, a randomized nonlinear component analysis based on PCA and canonical correlation analysis (CCA) has been introduced by Lopez-Paz [17] for data transformation. However, PCA shows poor performance on the blind source separation problem due to more complex dependencies among features [18], [19].…”

Section: Introductionmentioning

confidence: 99%

bSSA: Binary Salp Swarm Algorithm With Hybrid Data Transformation for Feature Selection

et al. 2021

View full text Add to dashboard Cite

“…As we consider the Euclidean distance in estimating the CCF, we transform the angular data on RA and DEC into some linear form by the help of the method proposed in Chattopadhyay, Mondal, & Biswas (2015). Let θ be the angular data with the unique mode φ, then its linear form is given by 1 − cos(θ − φ).…”

Section: Linearization Of Angular Datamentioning

confidence: 99%

Two phase formation of massive elliptical galaxies: study through cross-correlation including spatial effect

Modak

Chattopadhyay

2017

Astrophys Space Sci

Self Cite

View full text Add to dashboard Cite

Area of study is the formation mechanism of the present-day population of elliptical galaxies, in the context of hierarchical cosmological models accompanied by accretion and minor mergers. The present work investigates the formation and evolution of several components of the nearby massive early-type galaxies (ETGs) through cross-correlation function (CCF), using the spatial parameters right ascension (RA) and declination (DEC), and the intrinsic parameters mass (M * ) and size. According to the astrophysical terminology, here these variables, namely mass, size, RA and DEC are termed as parameters, whereas the unknown constants involved in the kernel function are called hyperparameters. Throughout this paper, the parameter size is used to represent the effective radius (R e ). Following Huang et al. (2013a), each nearby ETG is divided into three parts on the basis of its R e value. We study the CCF between each of these three components of nearby massive ETGs and the ETGs in the high redshift range, 0.5 < z ≤ 2.7. It is found that the innermost components of nearby ETGs are highly correlated with ETGs in the redshift range, 2 < z ≤ 2.7, known as 'red nuggets'. The intermediate and the outermost parts have moderate correlations with ETGs in the redshift range, 0.5 < z ≤ 0.75. The quantitative measures are highly consistent with the two phase formation scenario of nearby massive ETGs, as suggested by various authors, and resolve the conflict raised in a previous work (De, Chattopadhyay, & Chattopadhyay 2014) suggesting other possibilities for the formation of Soumita Modak

show abstract

“…Local and atmospheric sources of pollution may influence pollution concentrations, either directly, or indirectly, or both. Empirical and theoretical studies applied sophisticated strategies to link pollution with (observable and unobservable) atmospheric and other sources (see, e.g., Chattopadhyay et al, 2015;Cooley et al, 2012;Lagona et al, 2011;Paciorek et al, 2009;Park et al, 2001), modeling the fluctuations of concentrations over long periods, and accounting for uncertainty in the observed measurements. Although it is known that air pollution is a complex mixture of multiple pollutants, the pollution analyzes are in general conducted on each pollutant separately (see, e.g., Martinez-Zarzoso and Maruotti, 2013;Bornn et al, 2012;Greven et al, 2011;Shaddick et al, 2008).…”

Section: Introductionmentioning

confidence: 99%

Dynamic mixtures of factor analyzers to characterize multivariate air pollutant exposures

Maruotti¹,

Bulla²,

Lagona³

et al. 2017

Ann. Appl. Stat.

View full text Add to dashboard Cite

The assessment of pollution exposure is based on the analysis of multivariate time series that include the concentrations of several pollutants as well as the measurements of multiple atmospheric variables. It typically requires methods of dimensionality reduction that are capable to identify potentially dangerous combinations of pollutants and, simultaneously, to segment exposure periods according to air quality conditions. When the data are high-dimensional, however, efficient methods of dimensionality reduction are challenging because of the formidable structure of cross-correlations that arise from the dynamic interaction between weather conditions and natural/anthropogenic pollution sources. In order to assess pollution exposure in an urban area while taking the above mentioned difficulties into account, we develop a class of parsimonious hidden Markov models. In a multivariate time-series setting, this approach allows to simultaneously perform temporal segmentation and dimensionality reduction. We specifically approximate the distribution of multiple pollutant concentrations by mixtures of factor analysis models, whose parameters evolve according to a latent Markov chain. Covariates are included as predictors of the chain transition probabilities. Parameter constraints on the factorial component of the model are exploited to tune the flexibility of dimensionality reduction. In order to estimate the model parameters efficiently, we propose a novel three-step Alternating Expected Conditional Maximization (AECM) algorithm, which is also assessed in a simulation study. In the case study, the proposed methods were capable (1) to describe the exposure to pollution in terms of a few latent regimes, (2) to associate these regimes with specific combinations of pollutant concentration levels as well as distinct correlation structures between concentrations, and (3) to capture the influence of weather conditions on transitions between regimes.

show abstract

Independent component analysis and clustering for pollution data

Cited by 7 publications

References 12 publications

bSSA: Binary Salp Swarm Algorithm With Hybrid Data Transformation for Feature Selection

bSSA: Binary Salp Swarm Algorithm With Hybrid Data Transformation for Feature Selection

Two phase formation of massive elliptical galaxies: study through cross-correlation including spatial effect

Dynamic mixtures of factor analyzers to characterize multivariate air pollutant exposures

Contact Info

Product

Resources

About