Hidden Markov Model for Generalized Normal Distribution

[yanpanlau]

Hi,

     I am trying to create a Hidden Markov Model for Generalized Normal Distribution. 

But I got the error:

"Specified method is not supported". Could you please help?

       double[,] transition = 
        {
            {0.8, 0.2},
            {0.2, 0.8}
        };

        //Create the emission matrix B
        GeneralizedNormalDistribution[] emission = 
        {
            new GeneralizedNormalDistribution(0,1,1),
            new GeneralizedNormalDistribution(0,1,1),
        };

        // Create the initial probabilities pi 
        double[] initial =
        {
            0.6, 0.4
        };

        //Create a new Hidden Markov Model
        HiddenMarkovModel<GeneralizedNormalDistribution> model = new HiddenMarkovModel<GeneralizedNormalDistribution>(transition, emission, initial);

        BaumWelchLearning<GeneralizedNormalDistribution> teacher = new BaumWelchLearning<GeneralizedNormalDistribution>(model)
        {
            Tolerance = 1,
            Iterations = 0,
        };

        double[] sequence = new double[] { 0, 0, 0, 1, 1, 1, 2, 2, 3, 2, 3, 4, 6, 6, 6 };
        double likelihood = teacher.Run(sequence);

[blaisexen]

hi,
the way I see and tested here, it is not supported yet.

[cesarsouza]

Hi everyone,

Thanks @blaisexen you are correct - estimating GeneralizedNormalDistributions is not supported by the framework yet.

@yanpanlau thanks for opening the issue! As I had just mentioned, the problem is that GeneralizedNormalDistribution can not yet be estimated from data yet. I can try to add an estimation method for it in the next release, but please let me know if there is any estimation method that you would be mostly interested in.

Regards,
Cesar

[yanpanlau]

Thanks Cesar. Yes I am interested to do HMM with Generalized Normal Distribution, with the version 1(i.e. exponential power distribution). Thanks

…

On Sat, Oct 14, 2017 at 5:51 PM, César Souza ***@***.***> wrote: Hi everyone, Thanks @blaisexen <https://github.com/blaisexen> you are correct - estimating GeneralizedNormalDistributions is not supported by the framework yet. @yanpanlau <https://github.com/yanpanlau> thanks for opening the issue! As I had just mentioned, the problem is that GeneralizedNormalDistribution can not yet be estimated from data yet. I can try to add an estimation method for it in the next release, but please let me know if there is any estimation method that you would be mostly interested in. Regards, Cesar — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#945 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AO1sYwq00IfFLI5dtSIIIfzijM2VWie3ks5ssIQNgaJpZM4P1LNw> .