[mistake in docs] Deep learning trainingOp​tions.Vali​dationPati​ence is not Inf by default

Documentation :

 'ValidationPatience' — Patience of validation stopping
 Inf (default) | positive integer

Running this code:

trainingOptions('sgdm')

results in:

 Momentum: 0.9000
             InitialLearnRate: 0.0100
    LearnRateScheduleSettings: [1×1 struct]
             L2Regularization: 1.0000e-04
      GradientThresholdMethod: 'l2norm'
            GradientThreshold: Inf
                    MaxEpochs: 30
                MiniBatchSize: 128
                      Verbose: 1
             VerboseFrequency: 50
               ValidationData: []
          ValidationFrequency: 50
           ValidationPatience: 5
                      Shuffle: 'once'
               CheckpointPath: ''
         ExecutionEnvironment: 'auto'
                   WorkerLoad: []
                    OutputFcn: []
                        Plots: 'none'
               SequenceLength: 'longest'
         SequencePaddingValue: 0

Where ValidationPatience is 5...

Just letting you know...

Answer 1

[Tish Sheridan]

Thanks for confirming. The doc for your 18a release shows the old default to match the code you are seeing: https://www.mathworks.com/help/releases/R2018a/nnet/ref/trainingoptions.html#mw_c2ab36f1-691f-40d4-b894-cba9f7b27100 The doc for 18b shows the new default. The web doc always shows the current release. You can see the doc for your release by clicking Other Releases on the doc home page to get here: https://www.mathworks.com/help/doc-archives.html

Answer 2

[Tish Sheridan]

Hello - what version are you running? Starting in R2018b, the default 'ValidationPatience' changed to Inf. For more detail see under "Functionality being removed or changed" in the 18b release notes here: Release notes

[tesarj13]

I am using 2018a.