Software for Flexible Bayesian Modeling

Version of 1997-07-22. Copyright (c) 1995, 1996, 1997 by Radford M. Neal

This is an index to documenation for software implementing flexible Bayesian models based on neural networks, Gaussian processes, and finite and infinite mixture models, implemented using Markov chain Monte Carlo methods.

Introduction, overview, and examples:

Notes on the present release and past releases:

Alphabetical index within directories:

     UTIL              MC              NET             GP           MIX

   data-spec      mc                net             gp            mix
   extract        mc-quantities     net-dvar        gp-cov        mix-cases
   find-min       mc-spec           net-display     gp-display    mix-display
   grid           mc-temp-filter    net-eval        gp-eigen      mix-extensions
   log            mc-temp-sched     net-gd          gp-eval       mix-gen
   log-append     xxx-grad-test     net-gen         gp-gen        mix-mc
   log-copy       xxx-mc            net-hist        gp-mc         mix-quantities 
   log-last       xxx-stepsizes     net-mc          gp-pred       mix-spec
   log-records                      net-plt         gp-quantities  
   log-types                        net-pred        gp-spec
   model-spec                       net-quantities
   numin                            net-rej 
   prior                            net-spec 
   quantities                       
   rand-seed 
   series   
   xxx-hist
   xxx-plt
A more detailed index is given below. Note that documentation on some programs used only to test the software is not included in either index. The documentation for the demonstration of sampling from a bivariate Gaussian distribution [bvg] is also not indexed above, but is listed below.

The most important documentation files are marked below with an asterisk.

Generic utility programs [util]:

  * log             Facilities for handling log files
    log-types       Types of log file records used by various programs

    log-copy        Copy part of a log file to a new log file
    log-append      Append records from one log file to the end of another
    log-last        Display the index of the last record in a log file
    log-records     List all records in a log file

  * data-spec       Specify data sets for training and testing
  * numin           Facilities for input of numeric data

  * model-spec      Specify model for targets
  * prior           Meaning and syntax of prior specifications

    find-min        Find entry with minimum value (for cross validation)

    grid            Output a grid of points
    extract         Extract items at random from a data file

  * rand-seed       Specify a random number seed

  * quantities      Numeric quantities obtainable from log files

  * xxx-plt         Write quantities from a log file, suitable for plotting
    xxx-hist        Build a histogram for a quantity obtained from a log file

  * series          Analyse stationary time series data

Markov chain Monte Carlo facilities [mc]:

  * mc              Programs and modules supporting Markov chain Monte Carlo
  * mc-spec         Specify how to do the Markov chain simulation
  * xxx-mc          Run Markov chain simulation

  * mc-quantities   Quantities from log files relating to Monte Carlo

    mc-temp-sched   Specify temperature schedule for tempering methods
    mc-temp-filter  Copy only iterations at a given temperature

    xxx-grad-test   Test the correctness of the energy gradient computations
    xxx-stepsizes   Display and evaluate stepsizes used for dynamics

Neural network models [net]:

  * net             Bayesian inference for neural networks using MCMC
  * net-spec        Create a new network, or display existing specifications

  * net-mc          Do Markov chain simulation to sample networks
  * net-gen         Generate networks from the prior, or with fixed values
    net-gd          Train a network by gradient descent in the error

  * net-display     Print network parameters and/or hyperparameters

  * net-quantities  Quantities from log files relating to networks
  * net-plt         Get quantities from a net log file, suitable for plotting
    net-hist        Build histogram for quantity obtained from a net log file

  * net-pred        Make predictions for test cases

    net-eval        Evaluate network functions over a grid
    net-dvar        Find the variance of a difference in function values

    net-rej         Generate networks from the posterior by rejection sampling

Gaussian process models [gp]:

  * gp              Bayesian modelling using Gaussian processes
  * gp-spec         Specify a Gaussian process model, or display existing spec

  * gp-mc           Use Markov chain to sample Gaussian process hyperparameters
  * gp-gen          Generate GP hyperparameters randomly, or fix them

  * gp-display      Print Gaussian process hyperparameters & other information
  * gp-quantities   Quantities from log files relating to Gaussian processes

  * gp-pred         Make predictions for test cases using Gaussian process
    gp-eval         Evaluate function drawn from a Gaussian process over a grid

    gp-cov          Print covariance matrix for a Gaussian process 
    gp-eigen        Find eigenvalues/vectors of covariance matrix

Bayesian inference for mixture models [mix]:

  * mix             Bayesian inference for mixture models
  * mix-spec        Specify a mixture model, or display existing spec

  * mix-mc          Use Markov chain to do sampling for a mixture model
  * mix-gen         Generate hyperparameters randomly, or fix them

  * mix-display     Print mixture model parameters, hyperparameters, etc.
  * mix-quantities  Quantities from log files relating to mixture models

    mix-cases       Generate cases from a mixture model

    mix-extensions  Possible extensions to the mixture modeling software

Markov chain sampling for a bivariate Gaussian [bvg]:

    bvg             Demo of Markov chain sampling from a bivariate Gaussian
    bvg-spec        Specify a bivariate Gaussian distribution to sample from
    bvg-mc          Do Markov chain simulation for a bivariate Gaussian
    bvg-plt         Get quantities from a bvg log file, suitable to plot