MLP

MLP is available in Forecasting and Decision Trees.

The MLP procedure fits a particular kind of neural network called a multilayer perceptron. The multilayer perceptron uses a feedforward architecture and can have multiple hidden layers. It is one of the most commonly used neural network architectures.

MLP dependent variable [(MLEVEL = {S})] [dependent variable...]
                                  {O}
                                  {N}

    [BY factor list] [WITH covariate list]

[/EXCEPT VARIABLES = varlist]

[/RESCALE [COVARIATE = {STANDARDIZED**}]
                       {NORMALIZED    }
                       {ADJNORMALIZED }
                       {NONE          }

          [DEPENDENT = {STANDARDIZED                         }]]
                       {NORMALIZED    [(CORRECTION = {0.02**})]}
                                                     {number}
                       {ADJNORMALIZED [(CORRECTION = {0.02**})]}
                                                     {number}
                       {NONE                                 }

[/PARTITION {TRAINING = {70**   } TESTING = {30**   } HOLDOUT = {0**    }}]
                        {integer}           {integer}           {integer}
            {VARIABLE = varname                                       }

[/ARCHITECTURE [AUTOMATIC = {YES**} [(MINUNITS = {1**    }, MAXUNITS = {50**   })]]
                                                 {integer}             {integer}
                            {NO   }

               [HIDDENLAYERS = {1** [(NUMUNITS = {AUTO** })]       }]
                                                 {integer}
                               {2 [(NUMUNITS = {AUTO**          })]}
                                               {integer, integer}
               [HIDDENFUNCTION = {TANH** }]   [OUTPUTFUNCTION = {IDENTITY}]]
                                 {SIGMOID}                      {SIGMOID }
                                                                {SOFTMAX }
                                                                {TANH    }

[/CRITERIA [TRAINING = {BATCH**  }]   [MINIBATCHSIZE = {AUTO** }]
                       {ONLINE   }                     {integer}
                       {MINIBATCH}

           [MEMSIZE = {1000** }]      [OPTIMIZATION = {GRADIENTDESCENT}]
                      {integer}                       {SCALEDCONJUGATE}

           [LEARNINGINITIAL = {0.4** }]   [LEARNINGLOWER = {0.001**}]
                              {number}                     {number }

           [MOMENTUM = {0.9** }]   [LEARNINGEPOCHS = {10**   }]
                       {number}                      {integer}

           [LAMBDAINITIAL = {0.0000005**}]   [SIGMAINITIAL = {0.00005**}]
                            {number     }                    {number   }

           [INTERVALCENTER = {0**   }]    [INTERVALOFFSET = {0.5** }]]
                             {number}                       {number}

[/STOPPINGRULES [ERRORSTEPS = {1**    } [(DATA = {AUTO** })]]
                              {integer}          {BOTH   }

                [TRAININGTIMER = {ON**} [(MAXTIME = {15**  })]]
                                 {OFF }             {number}

                [MAXEPOCHS = {AUTO** }
                             {integer}]

                [ERRORCHANGE = {0.0001**}]    [ERRORRATIO = {0.001**}]]
                               {number  }                   {number }

[/MISSING USERMISSING = {EXCLUDE**}]
                        {INCLUDE  }

[/PRINT [CPS**]  [NETWORKINFO**]  [SUMMARY**]  [CLASSIFICATION**]
        [SOLUTION]  [IMPORTANCE]  [NONE]]

[/PLOT [NETWORK**]  [PREDICTED]  [RESIDUAL]  [ROC]
       [GAIN]  [LIFT]  [NONE]]

[/SAVE [PREDVAL[(varname [varname...])]]

       [PSEUDOPROB[(rootname[:{25     }] [rootname...])]]]
                            {integer}

[/OUTFILE [MODEL = 'file' ['file'...]]]

** Default if the subcommand or keyword is omitted.

This command reads the active dataset and causes execution of any pending commands. See the topic Command Order for more information.

Syntax for the MLP command can be generated from the Multilayer Perceptron dialog.

Release History

Release 16.0

• Command introduced.

Example

MLP dep_var BY A B C WITH X Y Z.