prtools.m

模式识别工具箱。非常丰富的底层函数和常见的统计识别工具

% Pattern Recognition Tools
% Version 4.1.3 10-Jun-2008
%
%Datasets and Mappings (just most important routines)
%---------------------
%dataset     Define dataset from datamatrix and labels
%datasets    List information on datasets (just help, no command)
%datafile    Define dataset from directory of object files 
%datafiles   List information on datafiles (just help, no command)
%classsizes  Retrieve sizes of classes
%gencirc     Generation of a one-class circular dataset
%genclass    Generate class frequency distribution
%genlab      Generate dataset labels
%getlab      Retrieve object labels from datasets and mappings
%getnlab     Retrieve nummeric object labels from dataset
%getfeat     Retrieve feature labels from datasets and mappings
%get         Get fields from datasets or mappings
%setdata     Change data in dataset
%setlabels   Change labels of dataset or mapping
%addlabels   Add additional labelling
%changelablist  Change current active labeling
%multi_labeling List information on multi-labeling (help only)
%mapping     Define and retrieve mapping and classifier from data
%mappings    List information on mappings (just help, no command)
%renumlab    Convert labels to numbers
%matchlab    Match different labelings
%primport    Convert old datasets to present PRTools definition
%remclass    Remove a class from a dataset
%seldat      Retrieve part of a dataset
%
%Data Generation (more in prdatasets)
%---------------
%circles3d   Create a dataset containing 2 circles in 3 dimensions
%lines5d     Create a dataset containing 3 lines in 5 dimensions
%gauss       Generation of multivariate Gaussian distributed data
%gendat      Generation of classes from given data set
%gendatb     Generation of banana shaped classes
%gendatc     Generation of circular classes
%gendatd     Generation of two difficult classes
%gendath     Generation of Highleyman classes
%gendati     Generation of random windows from images
%gendatk     Nearest neighbour data generation
%gendatl     Generation of Lithuanian classes
%gendatm     Generation of 8 2d classes
%gendatp     Parzen density data generation
%gendatr     Generate regression dataset from data and target values
%gendats     Generation of two Gaussian distributed classes
%gendatw     Sample dataset by given weigths
%gentrunk    Generation of Trunk's example
%prdata      Read data from file
%seldat      Select classes / features / objects from dataset
%getwindows  Get pixel feature vectors around given pixels in image dataset
%prdataset   Read existing dataset from file
%prdatasets  Overview of all datasets and data generators
%
%Datafiles
%---------
%datafile    Define datafile from set of files in directory
%savedatafile Save datafile, store intermediate result
%filtm       Mapping for arbitrary processing of datafile
%
%Linear and Quadratic Classifiers (*operate on datasets and datafiles)
%--------------------------------
%fisherc     Minimum least square linear classifier
%ldc         Normal densities based linear (muli-class) classifier
%loglc       Logistic linear classifier
%nmc         Nearest mean linear classifier
%nmsc        Scaled nearest mean linear classifier
%quadrc      Quadratic classifier
%qdc         Normal densities based quadratic (multi-class) classifier
%udc         Uncorrelated normal densities based quadratic classifier
%klldc       Linear classifier based on KL expansion of common cov matrix
%pcldc       Linear classifier based on PCA expansion on the joint data
%polyc       Add polynomial features and run arbitrary classifier
%subsc       Subspace classifier
% 
%classc      Converts a mapping into a classifier
%labeld      Find labels of objects by classification
%logdens     Convert density estimates to log-densities for more accuracy
%rejectc     Creates reject version of exisiting classifier
%testc       General error estimation routine for trained classifiers
%
%Other Classifiers 
%-----------------
%knnc        k-nearest neighbour classifier (find k, build classifier)
%knn_map     Map a dataset on a K-NN classifier, back end routine
%testk       Error estimation for k-nearest neighbour rule
%edicon      Edit and condense training sets
%
%weakc       Weak classifier
%stumpc      Decision stump classifier
%adaboostc   ADABoost classifier
%
%parzenc     Parzen classifier
%parzendc    Parzen density based classifier
%parzen_map  Map a dataset on a Parzen classfier, back end routine
%testp       Error estimation for Parzen classifier
%
%treec       Construct binary decision tree classifier
%tree_map    Map a dataset on a decision tree, back end routine
%naivebc     Naive Bayes classifier
%bpxnc       Feed forward neural network classifier by backpropagation
%lmnc        Feed forward neural network by Levenberg-Marquardt rule
%neurc       Automatic neural network classifier
%perlc       Linear perceptron 
%rbnc        Radial basis neural network classifier
%rnnc        Random neural network classifier
%ffnc        Feed-forward neural net classifier back-end routine
%
%fdsc        Feature based dissimilarity space classifier
%
%svc         Support vector classifier
%svo         Support vector optimizer
%nusvc       Support vector classifier
%nusvo       Support vector optimizer
%rbsvc       Radial basis SV classifier
%kernelc     General kernel/dissimilarity based classification
%
%Normal Density Based Classification
%-----------------------------------
%distmaha    Mahalanobis distance
%meancov     Estimation of means and covariance matrices from multiclass data
%nbayesc     Bayes classifier for given normal densities
%normal_map  Back-end routine for computing normal densities
%ldc         Normal densities based linear (muli-class) classifier
%qdc         Normal densities based quadratic (multi-class) classifier
%udc         Uncorrelated normal densities based quadratic classifier
%mogc        Mixture of gaussians classification
%testn       Error estimate of discriminant on normal distributions
%
%Feature Selection
%-----------------
%feateval    Evaluation of a feature set
%featrank    Ranking of individual feature permormances
%featsel     Feature Selection
%featselb    Backward feature selection
%featself    Forward feature selection
%featsellr   Plus-l-takeaway-r feature selection
%featseli    Feature selection on individual performance
%featselm    Feature selection map, general routine for feature selection
%featselo    Branch and bound feature selection
%featselp    Floating forward feature selection
%
%Classifiers and tests (general)
%-------------------------------
%bayesc      Bayes classifier by combining density estimates
%classim     Classify image using a given classifier
%classc      Convert mapping to classifier
%labeld      Find labels of objects by classification
%cleval      Classifier evaluation (learning curve)
%clevalb     Classifier evaluation (learning curve), bootstrap version
%clevalf     Classifier evaluation (feature size curve)
%clevals     Classifier evaluation (feature /learning curve), bootstrap
%confmat     Computation of confusion matrix
%costm       Cost mapping, classification using costs
%crossval    Crossvalidation 
%cnormc      Normalisation of classifiers
%disperror   Display error matrix with information on classifiers and datasets
%labelim     Construct image of labeled pixels
%logdens     Convert density estimates to log-densities for more accuracy
%mclassc     Computation of multi-class classifier from 2-class discriminants
%regoptc     Optimisation of regularisation and complexity parameters
%reject      Compute error-reject trade-off curve
%roc         Receiver-operator curve (ROC)
%testc       General error estimation routine for trained classifiers
%testauc     Estimate error as area under the ROC
%
%Mappings
%--------
%affine      Construct affine (linear) mapping from parameters
%bhatm       Two-class Bhattacharryya mapping
%cmapm       Compute some special maps
%featselm    Feature selection map, general routine for feature selection
%fisherm     Fisher mapping
%invsigm     Inverse sigmoid map
%filtm       Arbitrary operation on objects in datafiles/datasets
%gaussm      Mixture of Gaussians density estimation
%kernelm     Kernel mapping
%klm         Decorrelation and Karhunen Loeve mapping (PCA)
%klms        Scaled version of klm, useful for prewhitening
%knnm        k-Nearest neighbor density estimation
%mclassm     Computation of mapping from multi-class dataset
%map         General routine for computing and executing mappings
%nlfisherm   Nonlinear Fisher mapping
%normm       Object normalization map
%parzenm     Parzen density estimation