Gaussian distribution for ObservedValue RRS feed

  • Question

  • Hello everyone,

    I'm having trouble with implementing a simple Bayesian inference method. There is one parameter A, defined as a Guassian distribution. The observations of A provide information in the form of Gaussian distributions, A_observed_mean & A_observed_precision, which should be used to update the mean and precision  of A. Is it possible to provide observed information to the inference engine as a Gaussian distribution? It should look something like:

    A_observation.ObservedValue = GaussianFromMeanAndPrecision(A_observed_mean , A_observed_precision);

    I know the code isn't correct, but I hope the idea is clear enough to provide you with an idea of my problem. Thanks in advance.

    Best regards

    • Edited by Reinaert Friday, May 31, 2013 3:41 AM
    Friday, May 31, 2013 3:39 AM

All replies

  • Instead of thinking in terms of an inference procedure, try thinking in terms of a generative model.  That is, how are the observations generated from the unknown parameters? See the examples, such as Tutorial 3: Learning A Gaussian.   Infer.NET will work out the appropriate inference steps.
    Monday, June 3, 2013 2:38 PM