Variational Message Passing (VMP) Error RRS feed

  • Question

  • I have adapted the LearningAGaussian example by adding a Y variable,that is dependent on X (the variable of which the parameters need to be learned) and observing Y instead of X . The default inference type works, although I am not sure what type of inference this is, but VMP gives an error.

    Is there any reason why VMP should not work with this model?

    Also, what is the default inference method?

    The error is: 

    Error 0: This model is not supported with VariationalMessagePassing due to Factor.Plus(double Sum, double A, double B). Try using a different algorithm or expressing the model differently in

    Factor.Plus(vdouble11_use, Q_uses[0])
    Details: Variational Message Passing does not support a Plus factor with fixed output and two random inputs. using parameter types: [Sum] double,[A] Gaussian,[to_A] Gaussian,[B] Gaussian,[to_B] Gaussian,[result] Gaussian
    Tuesday, July 11, 2017 3:32 PM