Type: Package
Title: Sparse Generative Model and Its EM Algorithm
Version: 0.1-0
Date: 2015-09-05
Author: Charles Bouveyron, Julien Chiquet, Pierre Latouche, Pierre-Alexandre Mattei
Maintainer: Julien Chiquet <julien.chiquet@gmail.com>
Description: Implements a generative model that uses a spike-and-slab like prior distribution obtained by multiplying a deterministic binary vector. Such a model allows an EM algorithm, optimizing a type-II log-likelihood.
License: GPL-2 | GPL-3 [expanded from: GPL (≥ 2)]
Imports: methods
Repository: CRAN
Repository/R-Forge/Project: spinyreg
Repository/R-Forge/Revision: 11
Repository/R-Forge/DateTimeStamp: 2015-09-07 10:50:53
Date/Publication: 2015-09-07 18:18:03
NeedsCompilation: no
Packaged: 2015-09-07 11:07:11 UTC; rforge

spinyReg

Description

Computethe path of solution of a spinyReg fit.

Usage

spinyreg(X, Y, alpha = 0.1, gamma = 1, z = rep(1, ncol(X)),
  intercept = TRUE, normalize = TRUE, verbose = 1, recovery = TRUE,
  maxit = 1000, eps = 1e-10)

Arguments

X

matrix of features. Do NOT include intercept.

Y

matrix of responses.

alpha

numeric scalar; prior value for the alpha parameter (see the model's details). Default is 0.1.

gamma

numeric scalar; prior value for the gamma parameter (see the model's details). Default is 1.

z

numeric vector; prior support of active variable. Default is rep(1,p), meaning all variable activated

intercept

logical; indicates if a vector of intercepts should be included in the model. Default is TRUE.

normalize

logical; indicates if predictor variables should be normalized to have unit L2 norm before fitting. Default is TRUE.

verbose

integer; activate verbose mode from '0' (nothing) to '2' (detailed output).

should be included in the model. Default is TRUE.

recovery

logical; indicates if the full path of models should be inspected for model selection. Default is TRUE.

maxit

integer; the maximal number of iteration (i.e. number of alternated optimization between each parameter) in the Expectation/Maximization algorithm.

eps

a threshold for convergence. Default is 1e-10.

Value

an object with class spinyreg, see the documentation page spinyreg for details.

See Also

See also spinyreg.

Examples

## Not run: 
data <- read.table(file="http://statweb.stanford.edu/~tibs/ElemStatLearn/datasets/prostate.data")
x <- data[, 1:8]
y <- data[, 9]
out <- spinyreg(x,y,verbose=2)

## End(Not run)

Class "spinyreg"

Description

Class of object returned by the spinyreg function.

Slots

coefficients:

numeric vector of coefficients with respect to the original input. Contains the intercept if the model owns any.

alpha:

numeric scalar.

gamma:

numeric scalar.

normx:

Vector (class "numeric") containing the square root of the sum of squares of each column of the design matrix.

residuals:

Vector of residuals.

r.squared:

scalar giving the coefficient of determination.

fitted:

Vector of fitted values.

monitoring:

List (class "list") which contains various indicators dealing with the optimization process.

intercept:

Logical which indicates if a intercept is included in the model.

Methods

This class comes with the usual predict(object, newx, ...), fitted(object, ...), residuals(object, ...), coefficients(object, ...), print(object, ...) and show(object) generic (undocumented) methods.

mirror server hosted at Truenetwork, Russian Federation.