Octave Forge nnet

54 lines (40 with data), 2.4 kB

Summary of important user-visible changes for Neural Networks package 0.2.0:
----------------------------------------------------------------------------

 ** The following datasets have been added: house_dataset

 ** The functions ind2vec and vec2ind have been completely rewritten to accept
    a second argument as input and output respectively.

 ** The function dividerand has been completely rewritten to accept a matrix
    or a number as input.

 ** The following functions are new:

      minmax

 ** The following functions have been deprecated: 

      min_max (replaced by minmax)
      dhardlim.m
      dposlin.m
      dsatlin.m
      dsatlins.m

 ** The new function `minmax' meant to replace `min_max' can now handle single
    row matrices and cell arrays in a Matlab compatible way.

 ** The functions hardlim, hardlims, poslin, purelin, satlin, and satlins, have
    been completely rewritten to accept arrays (previously it only handled
    scalars correctly), and to always returns values of class double.

 ** The following functions have been made private:

      __analyzerows.m       __printBiasConnect.m     __printNumLayerDelays.m
      __calcjacobian.m      __printBiases.m          __printNumLayers.m
      __calcperf.m          __printB.m               __printNumOutputs.m
      __checknetstruct.m    __printInitFcn.m         __printNumTargets.m
      __copycoltopos1.m     __printInitParam.m       __printOutputConnect.m
      __dlogsig.m           __printInputConnect.m    __printOutputs.m
      __dpurelin.m          __printInputs.m          __printPerformFcn.m
      __dradbas.m           __printInputWeights.m    __printPerformParam.m
      __dtansig.m           __printIW.m              __printTargetConnect.m
      __getx.m              __printLayerConnect.m    __printTargets.m
      __init.m              __printLayers.m          __printTrainFcn.m
      __mae.m               __printLayerWeights.m    __printTrainParam.m
      __mse.m               __printLW.m              __randomisecols.m
      __newnetwork.m        __printMLPHeader.m       __rerangecolumns.m
      __optimizedatasets.m  __printNetworkType.m     __setx.m
      __printAdaptFcn.m     __printNumInputDelays.m  __trainlm.m
      __printAdaptParam.m   __printNumInputs.m

 ** Package is no longer automatically loaded.

 ** Package is now dependent on Octave 3.4 or later.