 Summary of important user-visible changes for statistics 1.7.0:
-------------------------------------------------------------------

 Important Notice: 1) `mad`, `mean`, `median`, `std`, `var` functions
                      shadow core Octave's respective functions prior
                      to Octave v9.1
                   2) incompatibility with the `nan` package

 New functions:
 ==============

 ** fcnnpredict

 ** fcnntrain

 ** fitcdiscr

 ** fitcgam

 ** fitcnet

 ** fitcsvm

 ** loadmodel

 ** signrank

 New classdefs:
 ==============

 ** ClassificationDiscriminant

 ** ClassificationGAM

 ** ClassificationNeuralNetwork

 ** ClassificationPartitionedModel

 ** ClassificationSVM

 ** CompactClassificationDiscriminant

 ** CompactClassificationGAM

 ** CompactClassificationNeuralNetwork

 ** CompactClassificationSVM

 Improvements:
 =============

 ** BinomialDistribution: fixed computations on truncated distribution
                          methods (github issue #128)

 ** ClassificationKNN: new methods and various bug fixes

 ** finv: fix erratic output for very large DF2 (savannah bug #66034)

 ** fitcknn: support for cross-validation

 ** NegativeBinomialDistribution: fixed computations on truncated distribution
                                  methods (github issue #128)

 ** plot method for probability distribution objects updated to support 'cdf'
    and 'probability' PlotType options (github issue #129)

 ** PoissonDistribution: fixed computations on truncated distribution
                         methods (github issue #128)


 Summary of important user-visible changes for statistics 1.7.1:
-------------------------------------------------------------------

 Important Notice: 1) `mad`, `mean`, `median`, `std`, `var` functions
                      shadow core Octave's respective functions prior
                      to Octave v9.1
                   2) incompatibility with the `nan` package

 Improvements:
 =============

 ** ClassificationPartitionedModel: add input validation for cvpartition
                                    object (github issue #160)

 ** ClassificationNeuralNetwork: use named variable for timing training
                                 (github issue #169)

 ** kstest: add support for probability distribution objects, fix previous
            regression related to failing test (github issues #164, #165, #167),
            add more BISTs and a DEMO

 ** RegressionGAM: add iteration limit to the private fitGAM function to avoid
                   infinite loop with Netlib BLAS/LAPACK (github issue #160)


 Summary of important user-visible changes for statistics 1.7.2:
-------------------------------------------------------------------

 Important Notice: 1) `mad`, `mean`, `median`, `std`, `var` functions
                      shadow core Octave's respective functions prior
                      to Octave v9.1
                   2) incompatibility with the `nan` package

 Improvements:
 =============

 ** fcnnpredict, fcnntrain: various bug fixes, allow MacOS users to enable
                            support for OpenMP (github issues #168, #171, #172)

 ** editDistance: use correct index types to avoid compiler warnings
                  (github issue #171)


 Summary of important user-visible changes for statistics 1.7.3:
-------------------------------------------------------------------

 Important Notice: 1) `mad`, `mean`, `median`, `std`, `var` functions
                      shadow core Octave's respective functions prior
                      to Octave v9.1
                   2) incompatibility with the `nan` package

 New functions:
 ==============

 ** glmval

 Improvements:
 =============

 ** glmfit: fixed output and updated functionality and compatibility


 Summary of important user-visible changes for statistics 1.7.4:
-------------------------------------------------------------------

 Important Notice: 1) `mad`, `mean`, `median`, `std`, `var` functions
                      shadow core Octave's respective functions prior
                      to Octave v9.1
                   2) incompatibility with the `nan` package

 Improvements:
 =============

 ** ClassificationDiscriminant, ClassificationKNN, ClassificationNeuralNetwork:
                    fix type of returning labels and ClassNames to be the same
                    as input Y, add custom display and subsref/subsassgn methods

 ** confusionchart: fix empty figures in demos (github issue #178)

 ** ConfusionMatrixChart: properly update cdata in figure (github issue #178)

 ** fitdist: fix occasionally failing BISTs (github issue #174)

 ** kmeans: fix empty figures in demos (github issue #179)

