GLFM | General Latent Feature Modeling for Heterogeneous data

Description of the functions available in the GLFM package

function hidden = GLFM.infer(data,varargin)

% Wrapper .m function to call .cpp MATLAB wrapper (simplifies call)
% Three possible calls:
%           hidden = GLFM.infer(data)
%           hidden = GLFM.infer(data,hidden)
%           hidden = GLFM.infer(data,hidden,params)
%
%   Inputs:
%       data: structure with all input data information
%           (*) data.X: N*D observation matrix (raw)
%           (*) data.C: 1*D string array with input data types
%       (*) mandatory
%
%       ------------- optional ---------------------
%       hidden: hidden structure to initialize inference algorithm
%             hidden.Z: feature assignment N*K matrix
%       params: structure with simulation parameters and hyperparameters
% Output:
%    hidden:  structure with latent variables learned by the model
%	          (same output as GLFM.infer function).

function [Xcomplete,hidden] = GLFM.complete(data,varargin)

% Function to complete a matrix that has missing values with their MAP solution
% Possible calls:
%           [Xcomplete,hidden]  = GLFM.complete(data)
%           [Xcomplete,hidden]  = GLFM.complete(data,hidden) % init hidden.Z externaly
%           [Xcomplete,hidden]  = GLFM.complete(data,[],params) % struc. with parameters
%           [Xcomplete,hidden]  = GLFM.complete(data,hidden,params)
%
% Inputs:
%    data: structure with all input data information
%          (*) data.X: NxD observation matrix (raw) with missings
%          (*) data.C: 1xD string array with input data types
%    (*) mandatory
%
%       ------------- optional ---------------------
%    hidden: hidden structure to initialize inference algorithm
%            hidden.Z: feature assignment N*K matrix
%    params: structure with simulation parameters and hyperparameters
% Output:
%    Xcomplete: NxD input matrix with imputed missing values
%    hidden:  structure with latent variables learned by the model
%	          (same output as GLFM.infer function).

function X_map = GLFM.computeMAP(C, Zp, hidden, params)

% Function to generate the MAP solution corresponding to patterns in Zp
% Inputs:
%   C: 1*D string with data types, D = number of dimensions
%   Zp: P * K matrix of feature activation for which to compute the MAP estimate
%       (P is the number of obs.)
%   hidden: structure with latent variables learned by the model
%   params: structure of simulation parameters and hyperparameters
%
% Outputs:
%   X_map: P*D matrix with MAP estimate

function [xd, pdf] = GLFM.computePDF(data, Zp, hidden, params, d)

% Function to generate the PDF solutions corresponding to patterns in
% Zp, and dimension d
% Inputs:
%   data: data structure
%   Zp: P * K matrix of patterns (P is the number of patterns)
%   hidden: structure with latent variables learned by the model
%   params: structure of simulation parameters and hyperparameters
%   patterns: numP*K list of patterns to plot
%
% Outputs:
%   xd: 1*numS where numS is the number of points to compute
%   pdf: P*numS where P is the number of patterns to consider

function [] = GLFM.plotPatterns(data, hidden, params, Zp, varargin)

% Function to plot the inferred distribution and empirical histogram for each dimension of the observations.
% Inputs:
%   data: data structure
%   hidden: structure with latent variables learned by the model
%   params: structure of simulation parameters and hyperparameters
%   Zp: P * K matrix of patterns (P is the number of patterns)
%   ------ (optional) ------
%   colors: list of colors to plot
%   styles: list of styles for each line (for plot, not bar)
%   leg: legend to use (by default, use patterns as legend)
%   idxD: array of dimensions to plot
%
% Outputs:
%   void