R/CRF-package.R

In CRF: Conditional Random Fields

#' CRF - Conditional Random Fields
#' 
#' Library of Conditional Random Fields model
#' 
#' CRF is R package for various computational tasks of conditional random
#' fields as well as other probabilistic undirected graphical models of
#' discrete data with pairwise and unary potentials. The
#' decoding/inference/sampling tasks are implemented for general discrete
#' undirected graphical models with pairwise potentials. The training task is
#' less general, focusing on conditional random fields with log-linear
#' potentials and a fixed structure. The code is written entirely in R and C++.
#' The initial version is ported from UGM written by Mark Schmidt.
#' 
#' Decoding: Computing the most likely configuration 
#' \itemize{
#'   \item \code{\link{decode.exact}} Exact decoding for small graphs with brute-force search
#'   \item \code{\link{decode.chain}} Exact decoding for chain-structured graphs with the Viterbi algorithm 
#'   \item \code{\link{decode.tree}} Exact decoding for tree- and forest-structured graphs with max-product belief propagation 
#'   \item \code{\link{decode.conditional}} Conditional decoding (takes another decoding method as input) 
#'   \item \code{\link{decode.cutset}} Exact decoding for graphs with a small cutset using cutset conditioning
#'   \item \code{\link{decode.junction}} Exact decoding for low-treewidth graphs using junction trees 
#'   \item \code{\link{decode.sample}} Approximate decoding using sampling (takes a sampling method as input) 
#'   \item \code{\link{decode.marginal}} Approximate decoding using inference (takes an inference method as input) 
#'   \item \code{\link{decode.lbp}} Approximate decoding using max-product loopy belief propagation 
#'   \item \code{\link{decode.trbp}} Approximate decoding using max-product tree-reweighted belief propagtion 
#'   \item \code{\link{decode.greedy}} Approximate decoding with greedy algorithm 
#'   \item \code{\link{decode.icm}} Approximate decoding with the iterated conditional modes algorithm 
#'   \item \code{\link{decode.block}} Approximate decoding with the block iterated conditional modes algorithm 
#'   \item \code{\link{decode.ilp}} Exact decoding with an integer linear programming formulation and approximate using LP relaxation 
#' }
#' 
#' Inference: Computing the partition function and marginal probabilities
#' \itemize{ 
#'   \item \code{\link{infer.exact}} Exact inference for small graphs with brute-force counting 
#'   \item \code{\link{infer.chain}} Exact inference for chain-structured graphs with the forward-backward algorithm 
#'   \item \code{\link{infer.tree}} Exact inference for tree- and forest-structured graphs with sum-product belief propagation 
#'   \item \code{\link{infer.conditional}} Conditional inference (takes another inference method as input) 
#'   \item \code{\link{infer.cutset}} Exact inference for graphs with a small cutset using cutset conditioning 
#'   \item \code{\link{infer.junction}} Exact decoding for low-treewidth graphs using junction trees 
#'   \item \code{\link{infer.sample}} Approximate inference using sampling (takes a sampling method as input) 
#'   \item \code{\link{infer.lbp}} Approximate inference using sum-product loopy belief propagation 
#'   \item \code{\link{infer.trbp}} Approximate inference using sum-product tree-reweighted belief propagation 
#' }
#' 
#' Sampling: Generating samples from the distribution 
#' \itemize{ 
#'   \item \code{\link{sample.exact}} Exact sampling for small graphs with brute-force inverse cumulative distribution 
#'   \item \code{\link{sample.chain}} Exact sampling for chain-structured graphs with the forward-filter backward-sample algorithm 
#'   \item \code{\link{sample.tree}} Exact sampling for tree- and forest-structured graphs with sum-product belief propagation and backward-sampling 
#'   \item \code{\link{sample.conditional}} Conditional sampling (takes another sampling method as input) 
#'   \item \code{\link{sample.cutset}} Exact sampling for graphs with a small cutset using cutset conditioning 
#'   \item \code{\link{sample.junction}} Exact sampling for low-treewidth graphs using junction trees 
#'   \item \code{\link{sample.gibbs}} Approximate sampling using a single-site Gibbs sampler 
#' }
#' 
#' Training: Given data, computing the most likely estimates of the parameters
#' \itemize{
#'   \item \code{\link{train.crf}} Train CRF model
#'   \item \code{\link{train.mrf}} Train MRF model
#' }
#' 
#' Tools: Tools for building and manipulating CRF data
#' \itemize{
#'   \item \code{\link{make.crf}} Generate CRF from the adjacent matrix
#'   \item \code{\link{make.features}} Make the data structure of CRF features
#'   \item \code{\link{make.par}} Make the data structure of CRF parameters
#'   \item \code{\link{duplicate.crf}} Duplicate an existing CRF
#'   \item \code{\link{clamp.crf}} Generate clamped CRF by fixing the states of some nodes
#'   \item \code{\link{clamp.reset}} Reset clamped CRF by changing the states of clamped nodes
#'   \item \code{\link{sub.crf}} Generate sub CRF by selecting some nodes
#'   \item \code{\link{mrf.update}} Update node and edge potentials of MRF model
#'   \item \code{\link{crf.update}} Update node and edge potentials of CRF model
#' }
#' 
#' @name CRF-package
#' @aliases CRF-package CRF
#' @docType package
#' @keywords package
#' @author Ling-Yun Wu \email{wulingyun@@gmail.com}
#' 
#' @references J. Lafferty, A. McCallum, and F. Pereira. Conditional random fields:
#' Probabilistic models for segmenting and labeling sequence data. In \emph{the 
#' proceedings of International Conference on Machine Learning (ICML)}, pp. 282-289, 2001.  
#' @references Mark Schmidt. UGM: A Matlab toolbox for probabilistic undirected graphical models.
#' \url{http://www.cs.ubc.ca/~schmidtm/Software/UGM.html}, 2007.
#' 
#' @examples
#' 
#' library(CRF)
#' data(Small)
#' decode.exact(Small$crf)
#' infer.exact(Small$crf)
#' sample.exact(Small$crf, 100)
#' 
#' @useDynLib CRF, .registration = TRUE
#' 
NULL



#' Small CRF example
#' 
#' This data set gives a small CRF example
#' 
#' @format A list containing two elements:
#' \itemize{
#'   \item \code{crf} The CRF
#'   \item \code{answer} A list of 4 elements:
#'   \itemize{
#'     \item \code{decode} The most likely configuration
#'     \item \code{node.bel} The node belief
#'     \item \code{edge.bel} The edge belief
#'     \item \code{logZ} The logarithmic value of CRF normalization factor Z
#'   }
#' }
#' 
#' @name Small
#' @aliases Small
#' @docType data
#' @keywords datasets
#' @usage data(Small)
#' 
NULL



#' Chain CRF example
#' 
#' This data set gives a chain CRF example
#' 
#' @format A list containing two elements:
#' \itemize{
#'   \item \code{crf} The CRF
#'   \item \code{answer} A list of 4 elements:
#'   \itemize{
#'     \item \code{decode} The most likely configuration
#'     \item \code{node.bel} The node belief
#'     \item \code{edge.bel} The edge belief
#'     \item \code{logZ} The logarithmic value of CRF normalization factor Z
#'   }
#' }
#' 
#' @name Chain
#' @aliases Chain
#' @docType data
#' @keywords datasets
#' @usage data(Chain)
#' 
NULL



#' Tree CRF example
#' 
#' This data set gives a tree CRF example
#' 
#' @format A list containing two elements:
#' \itemize{
#'   \item \code{crf} The CRF
#'   \item \code{answer} A list of 4 elements:
#'   \itemize{
#'     \item \code{decode} The most likely configuration
#'     \item \code{node.bel} The node belief
#'     \item \code{edge.bel} The edge belief
#'     \item \code{logZ} The logarithmic value of CRF normalization factor Z
#'   }
#' }
#' 
#' @name Tree
#' @aliases Tree
#' @docType data
#' @keywords datasets
#' @usage data(Tree)
#' 
NULL



#' Loop CRF example
#' 
#' This data set gives a loop CRF example
#' 
#' @format A list containing two elements:
#' \itemize{
#'   \item \code{crf} The CRF
#'   \item \code{answer} A list of 4 elements:
#'   \itemize{
#'     \item \code{decode} The most likely configuration
#'     \item \code{node.bel} The node belief
#'     \item \code{edge.bel} The edge belief
#'     \item \code{logZ} The logarithmic value of CRF normalization factor Z
#'   }
#' }
#' 
#' @name Loop
#' @aliases Loop
#' @docType data
#' @keywords datasets
#' @usage data(Loop)
#' 
NULL



#' Clique CRF example
#' 
#' This data set gives a clique CRF example
#' 
#' @format A list containing two elements:
#' \itemize{
#'   \item \code{crf} The CRF
#'   \item \code{answer} A list of 4 elements:
#'   \itemize{
#'     \item \code{decode} The most likely configuration
#'     \item \code{node.bel} The node belief
#'     \item \code{edge.bel} The edge belief
#'     \item \code{logZ} The logarithmic value of CRF normalization factor Z
#'   }
#' }
#' 
#' @name Clique
#' @aliases Clique
#' @docType data
#' @keywords datasets
#' @usage data(Clique)
#' 
NULL



#' Rain data
#' 
#' This data set gives an example of rain data used to train CRF and MRF models
#' 
#' @format A list containing two elements:
#' \itemize{
#'   \item \code{rain} A matrix of 28 columns containing raining data (1: rain, 2: sunny).
#'     Each row is an instance of 28 days for one month.
#'   \item \code{months} A vector containing the months of each instance.
#' }
#' 
#' @name Rain
#' @aliases Rain
#' @docType data
#' @keywords datasets
#' @usage data(Rain)
#' 
#' @references Mark Schmidt. UGM: A Matlab toolbox for probabilistic undirected graphical models.
#' \url{http://www.cs.ubc.ca/~schmidtm/Software/UGM.html}, 2007.
#' 
NULL