Projection Pursuit for Cluster Identification (PPCI)
Implements recently developed projection pursuit algorithms for finding optimal linear cluster separators. The clustering algorithms use optimal hyperplane separators based on minimum density, Pavlidis et. al (2016) <https://…/15-307.pdf>; minimum normalised cut, Hofmeyr (2017) <doi:10.1109/TPAMI.2016.2609929>; and maximum variance ratio clusterability, Hofmeyr and Pavlidis (2015) <doi:10.1109/SSCI.2015.116>.

Summarize Text by Ranking Sentences (textrank)
The ‘textrank’ algorithm is an extension of the ‘Pagerank’ algorithm for text. The algorithm allows to summarize text by calculating how sentences are related to one another. This is done by looking at overlapping terminology used in sentences in order to set up links between sentences. The resulting sentence network is next plugged into the ‘Pagerank’ algorithm which identifies the most important sentences in your text and ranks them. More information can be found in the paper from Mihalcea, Rada & Tarau, Paul (2004) <http://…/W04-3252>.

Simultaneous Truth and Performance Level Estimation (stapler)
An implementation of Simultaneous Truth and Performance Level Estimation (STAPLE) <doi:10.1109/TMI.2004.828354>. This method is used when there are multiple raters for an object, typically an image, and this method fuses these ratings into one rating. It uses an expectation-maximization method to estimate this rating and the individual specificity/sensitivity for each rater.

Bayesian Analysis, No Gibbs (bang)
Provides functions for the Bayesian analysis of some simple commonly-used models, without using Markov Chain Monte Carlo (MCMC) methods such as Gibbs sampling. The ‘rust’ package <https://…/package=rust> is used to simulate a random sample from the required posterior distribution. At the moment three conjugate hierarchical models are available: beta-binomial, gamma-Poisson and a 1-way analysis of variance (ANOVA).

Adaptive Shrinkage of Correlation Vectors and Matrices (CorShrink)
Performs adaptive shrinkage of correlation and covariance matrices using a mixture model prior over the Fisher z-transformation of the correlations, Stephens (2016) <doi:10.1093/biostatistics/kxw041> with the method flexible in choosing a separate shrinkage intensity for each cell of the correlation or covariance matrices: it is particularly efficient in handling missing data in the data matrix.

Time-Varying Coefficients Linear Regression for Single and Multiple Equations (tvReg)
Fitting simultaneous equations with time varying coefficients, both for the case of independent equations and for the case of correlated equations.

Advertisements