Menu ▾ ▴

migration from 2.6 to 2.7 and 2.8

Open Discussion
guillaume T
2014-05-10
2014-07-20

  • guillaume T

    guillaume T - 2014-05-10

    I have been using neuroph 2.5.1 for a few years with some interesting results.
    I apparently can successfully upgrade to 2.6.
    I have also tried the latest 2.7 and 2.8. The problem is that from version 2.7 neuroph gives me a flat line as output.
    I am using the default set-up for the MultiLayerPerceptron.
    In order to migrate from 2.5 to 2.8 I had to replace
    TrainingSet with DataSet and SupervisedTrainingElement with DataSetRow.
    Everything seemed to go fine until I run the training and calculation which will now constantly give me a flat line output around 0.68/0.69
    Do I have to do more adjustments?

     

  • Zoran Sevarac

    Zoran Sevarac - 2014-05-14

    There shouldnt be any more adjustments. Can you provide a sample dataset to check? Are you training from code or GUI?

     

  • guillaume T

    guillaume T - 2014-05-14

    Hi Zoran,
    Thank you for your response.

    The network is a MultiLayerPerceptron with the following settings :
    Training topology : 3,7,1 - MaxIterations :1000 - MaxError :1.0E-4 - LearningRate 0.7

    I train using
    neuralNetwork.learn(trainingSet)
    And calculate using
    neuralNetwork.setInput(input);
    neuralNetwork.calculate();

    Here is a sample of data that would give me some rather consistent results using 2.5.1 as 2.6 but a flat line with versions above.

    Training input and desired
    Intputs,[1.0, 0.5, 0.0], Output[0.785606821734486]
    Intputs,[1.0, 0.5, 0.0], Output[0.5000620713972663]
    Intputs,[1.0, 0.0, 0.5], Output[0.5131025160333476]
    Intputs,[0.5, 0.0, 1.0], Output[0.4995641418857696]
    Intputs,[1.0, 0.0, 0.5], Output[0.6295420187846059]
    Intputs,[1.0, 0.5, 0.0], Output[0.4847516561166242]
    Intputs,[1.0, 0.5, 0.0], Output[0.5804892254860108]
    Intputs,[1.0, 0.5, 0.0], Output[0.4782081020761094]
    Intputs,[0.5, 0.0, 1.0], Output[0.5736435728565594]
    Intputs,[1.0, 0.0, 0.5], Output[0.4757771426279725]
    Intputs,[0.0, 0.5, 1.0], Output[0.53865006410737]
    Intputs,[0.0, 1.0, 0.5], Output[0.2938346910218203]
    Intputs,[1.0, 0.0, 0.5], Output[0.5728941074890153]
    Intputs,[0.5, 0.0, 1.0], Output[0.44013485464426816]
    Intputs,[1.0, 0.5, 0.0], Output[0.5372426431741694]
    Intputs,[1.0, 0.0, 0.5], Output[0.37939080392011393]
    Intputs,[0.0, 1.0, 0.5], Output[0.5282983644135463]
    Intputs,[1.0, 0.5, 0.0], Output[0.35835338495081587]
    Intputs,[0.0, 0.5, 1.0], Output[0.5653353797656009]
    Intputs,[1.0, 0.0, 0.5], Output[0.3032167032332047]
    Intputs,[0.0, 1.0, 0.5], Output[0.7561783032837601]
    Intputs,[0.0, 1.0, 0.5], Output[0.42256442104474906]
    Intputs,[1.0, 0.0, 0.5], Output[0.7196652938196935]
    Intputs,[0.5, 1.0, 0.0], Output[0.44018314540365144]
    Intputs,[1.0, 0.0, 0.5], Output[0.7040036771373075]
    Intputs,[0.0, 0.5, 1.0], Output[0.4025136247274861]
    Intputs,[1.0, 0.5, 0.0], Output[0.5227976836751145]
    Intputs,[0.5, 1.0, 0.0], Output[0.44067033615604234]
    Intputs,[0.5, 1.0, 0.0], Output[0.5412414324703206]
    Intputs,[0.5, 1.0, 0.0], Output[0.47676288489048335]
    Intputs,[1.0, 0.0, 0.5], Output[0.5419912481910825]
    Intputs,[0.0, 1.0, 0.5], Output[0.0]
    Intputs,[0.0, 1.0, 0.5], Output[0.6658509258662154]
    Intputs,[0.0, 1.0, 0.5], Output[0.10704069893996937]
    Intputs,[0.5, 1.0, 0.0], Output[0.5323921738014412]
    Intputs,[1.0, 0.5, 0.0], Output[0.49975298003304136]
    Intputs,[0.5, 1.0, 0.0], Output[1.0]

    Calculation input sample
    Caclulation input,[0.0, 0.5, 1.0]
    Caclulation input,[0.0, 0.5, 1.0]
    Caclulation input,[0.0, 0.5, 1.0]
    Caclulation input,[0.0, 0.5, 1.0]
    Caclulation input,[0.0, 0.5, 1.0]
    Caclulation input,[0.0, 0.5, 1.0]
    Caclulation input,[0.0, 0.5, 1.0]
    Caclulation input,[0.0, 0.5, 1.0]
    Caclulation input,[0.0, 0.5, 1.0]
    Caclulation input,[0.0, 0.5, 1.0]
    Caclulation input,[0.0, 0.5, 1.0]
    Caclulation input,[0.0, 0.5, 1.0]
    Caclulation input,[0.0, 0.5, 1.0]
    Caclulation input,[0.0, 0.5, 1.0]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.5, 1.0, 0.0]
    Caclulation input,[0.5, 1.0, 0.0]
    Caclulation input,[0.5, 1.0, 0.0]
    Caclulation input,[0.5, 1.0, 0.0]
    Caclulation input,[0.5, 1.0, 0.0]
    Caclulation input,[0.5, 1.0, 0.0]
    Caclulation input,[0.5, 1.0, 0.0]
    Caclulation input,[1.0, 0.5, 0.0]
    Caclulation input,[1.0, 0.5, 0.0]
    Caclulation input,[1.0, 0.5, 0.0]
    Caclulation input,[1.0, 0.5, 0.0]
    Caclulation input,[1.0, 0.5, 0.0]
    Caclulation input,[1.0, 0.5, 0.0]
    Caclulation input,[1.0, 0.5, 0.0]
    Caclulation input,[0.5, 1.0, 0.0]
    Caclulation input,[0.5, 1.0, 0.0]
    Caclulation input,[0.5, 1.0, 0.0]
    Caclulation input,[0.5, 1.0, 0.0]
    Caclulation input,[0.5, 1.0, 0.0]
    Caclulation input,[0.5, 1.0, 0.0]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]
    Caclulation input,[0.0, 1.0, 0.5]

    I am doing something wrong here?

     

    Last edit: guillaume T 2014-05-15

  • guillaume T

    guillaume T - 2014-07-20

    Hi Zoran,
    Thanks for the good work.
    Did you have a chance to look into this?

     

Log in to post a comment.