Sign in
Microsoft.com
 
Microsoft
 
 
 

locked
Approximative EM / approximate evidence maximization (Migrated from community.research.microsoft.com) RRS feed

 > 

  • Question

  • Question
    Sign in to vote
    0
    Sign in to vote

    TimSal posted on 05-31-2010 10:01 AM

    Hi guys,

    most interesting models I can think of require the determination of one or more 'hyper-parameters' that can not easily be infered using EP/variational methods. Does Infer.NET provide any automated way of estimating these parameters by performing approximate expectation maximization or by directly maximizing the approximate evidence?

    Cheers,

    Tim

    Friday, June 3, 2011 5:42 PM

Answers

  • Question
    Sign in to vote
    0
    Sign in to vote

    minka replied on 06-06-2010 4:42 AM

    By the way, we are continually increasing the set of parameters that can be inferred, and if you tell us what hyperparameters you are trying to infer, we can try to make these inferrable in the next version.

    Friday, June 3, 2011 5:43 PM

All replies

  • Question
    Sign in to vote
    0
    Sign in to vote

    minka replied on 06-04-2010 4:20 PM

    Infer.NET does not provide this automatically.  You would have to write your own code to optimize the hyperparameters, though you can use Infer.NET to compute any marginals required by the optimizer.

    Friday, June 3, 2011 5:43 PM
  • Question
    Sign in to vote
    0
    Sign in to vote

    minka replied on 06-06-2010 4:42 AM

    By the way, we are continually increasing the set of parameters that can be inferred, and if you tell us what hyperparameters you are trying to infer, we can try to make these inferrable in the next version.

    Friday, June 3, 2011 5:43 PM