mixpp: bdmtoolbox - List of available basic objects

bdmtoolbox - List of available basic objects

Basic objects of BDM are:
  • RV decriptor of data vectors and random variables
  • fnc functions of vector arguments
  • pdf pdf with non-empty conditioning part
  • epdf pdf with empty conditioning part
  • BM Bayesian model - (approximate) calculation of specific type of the Bayes rule,
  • DS Data Sources - recursive sources of data
  • logger object storing results of all experiments

See, BDM Use - Probability density functions and BDM Use - System, Data, Simulation for their introduction.

Working with RV

In bdmtoolbox, RV is represented by a matlab structure. It is recommended to use the following function to work with it:

  • RV, rva=RV('a',1), rvab=RV({'b','c'},[1,2]) create empty RV, RV named "a" of size 1, and vector of RV "b" and "c" of sizes 1 and 2, respectively.
  • RVtimes(rva,-1) set time (delay) of rva to -1 from current time.
  • RVjoin([rva,rvb]) joins rva and rvb into a vector

Basic functions

Predefined functions are:
  • constfn Class representing function , here rv is empty
  • linfn Class representing function
  • mexFnc Matlab extension of a function, calling predefined matlab functions
  • grid_fnc Function defined by values on a fixed grid and interpolated inbetween them

Square root decompositions of symetric matrices

For information purpose, these matrices are used in epdfs:
  • chmat Symmetric matrix stored in square root decomposition using upper cholesky
  • fsqmat Fake sqmat. This class maps sqmat operations to operations on full matrix
  • ldmat Matrix stored in LD form, (commonly known as UD)

Basic epdfs accessible from matlab are:

  • dirac Dirac delta function distribution
  • eDirich Dirichlet posterior density
  • eEF General conjugate exponential family posterior density
  • eEmp Weighted empirical density
  • egamma Gamma posterior density
  • egiw Gauss-inverse-Wishart density stored in LD form
  • egrid Discrete density defined on a continuous grid
  • eigamma Inverse-Gamma posterior density
  • eiWishartCh Inverse Wishart on Choleski decomposition
  • enorm<ldmat> Gaussian density with positive definite (decomposed) covariance matrix
  • enorm<chmat> Gaussian density with positive definite (decomposed) covariance matrix
  • enorm<fsqmat> Gaussian density with positive definite (decomposed) covariance matrix
  • elognorm LogNormal density
  • emix Mixture of epdfs
  • eprod Product of densities, elements may be unconditional, however, the result should be unconditioned pdf.
  • estudent< ldmat> Student-t density
  • estudent< ldmat> Student-t density
  • estudent< ldmat> Student-t density
  • euni Uniform density on rectangular support
  • eWishartCh Wishart in choleski decomposition

Basic (conditioned) pdfs accessible from matlab are:

  • mDirich Dirichlet random walk
  • mexBM BM with functions implemented in matlab
  • mexEpdf Epdf with functions implemented in matlab
  • mgamma Gamma random walk
  • mgamma_fix Gamma random walk around a fixed point
  • mgdirac Dirac delta pdf with geenral function of the support point
  • mgnorm< ldmat > Gaussian Pdf with general function for mean value
  • mgnorm< chmat > Gaussian Pdf with general function for mean value
  • mgnorm< fsqmat > Gaussian Pdf with general function for mean value
  • mguni Uniform density with geenral function of mean value
  • migamma Inverse-Gamma random walk
  • migamma_ref Inverse-Gamma random walk around a fixed point
  • mlnorm< ldmat> Normal distribution with linear function with linear function of mean value;
  • mlnorm< chmat> Normal distribution with linear function with linear function of mean value;
  • mlnorm< fsqmat> Normal distribution with linear function with linear function of mean value;
  • mlognorm Log-Normal random walk
  • mlstudent Student distributed linear function with linear function of mean value;
  • mmix Mixture of pdfs with constant weights, all pdfs are of equal RV and RVC
  • mprod Chain rule decomposition of epdf
  • mratio Class representing ratio of two densities which arise e.g. by applying the Bayes rule.
  • rwiWishartCh Random Walk on inverse Wishart

DataSources

  • MemDS Memory storage of off-line data column-wise
  • CsvFileDS CSV file data storage The constructor creates Data matrix from the records in a CSV file fname. The orientation can be of two types: 1. BY_COL which is default - the data are stored in columns; one column per time , one row per data item. 2. BY_ROW if the data are stored the classical CSV style. Then each column stores the values for data item, for ex. , one row for each discrete time instant
  • EpdfDS Simulate data from a static pdf (epdf)
  • PdfDS Simulate data from conditional density Still having only one density but allowing conditioning on either input or delayed values

Loggers - storage of results

  • mexlog Logger storing results into an mxArray
  • dirfilelog Logging into dirfile with buffer in memory
  • stdlog Simple logger used for debugging All data records are written out to std from where they could be send to file
  • ITppFileDS Read Data Matrix from an IT file

Mergers

  • merger_base Base class for general combination of pdfs on discrete support
  • merger_mix Merger using importance sampling with mixture proposal density

Bayesian Models

Basic filters:

  • ARX Linear Autoregressive model with Gaussian noise
  • ARXfrg ARX with conditioned on forgetting factor
  • ARXg ARX with Non-linear transformation + Gaussian noise
  • EKFCh Extended Kalman Filter in Square root
  • EKFfull Extended Kalman Filter in full matrices
  • KalmanCh Kalman filter in square root form
  • KalmanFull Basic Kalman filter with full matrices
  • MixEF Estimator of Mixtures of Exponential Family Densities
  • multiBM Estimator for Multinomial density
  • MultiModel (Switching) Multiple Models. The model runs several BMs in parallel and evaluates thier weights (fittness)
  • PF Particle filtering: Wrapper for particles
    • BootstrapParticle Class used in PF
    • MarginalizedParticle Class used in PF

Controllers (and related classes)

  • LQG Basic class computing LQG control
  • LQG_ARX Controller using ARX model for estimation and LQG designer for control
  • StateFromARX conversion function
  • StateSpace< ldmat> Basic elements of linear state-space model
  • StateSpace< chmat> Basic elements of linear state-space model

Multiple Patricipant Decision Makers

ARXAgent ARX agent
Generated on 2 Dec 2013 for mixpp by  doxygen 1.4.7