Sign in

Microsoft.com

Microsoft

Remove From My Forums

How to use posterior for testing in the LDA model (Migrated from community.research.microsoft.com)

Question
0

Sign in to vote

shakey posted on 11-21-2009 7:42 AM

I want to test the lda model on new documents, I take the method on the baysian point machine example:

setting the phi in the test model   to the posterior phi from the training model. However, I always get errors indicating violation of type constrain. Can anyone give me some help on this ? It seems SharedVariable is another kind of solution, but I cannot get it right either.

part of the code:

post_phi=engine.Infer[DistributionRefArray[Dirichlet,Vector]](phi)
test_post_phi=Variable[Vector].Random[DistributionRefArray[Dirichlet,Vector]](post_phi )

with (Variable.ForEach(test_CorpusSize)) as forBlock :

    with (Variable.ForEach(test_DocSize)) as forBlock :

        test_Z[test_CorpusSize][test_DocSize] = Variable.Discrete(test_theta[test_CorpusSize]).Attrib( ValueRange(TopicsNum))

        with (Variable.Switch(test_Z[test_CorpusSize][test_DocSize])) as swBlock :

            test_W[test_CorpusSize][test_DocSize] = Variable.Discrete(test_post_phi[ test_Z[test_CorpusSize][ test_DocSize] ])

Friday, June 3, 2011 5:26 PM

Answers

0

Sign in to vote
shakey replied on 11-21-2009 11:52 PM

Thanks for the reply.

I tried as you said using the SetTo method ,but the VariableArray[Vector].SetTo in IronPython does't seems to accept params other than the VariableArray[Vector]

type,I tried many times but failed. However, I find another way around:

post_phi=engine.Infer(phi)
test_post_phi = Variable.Array[Vector](TopicsNum).Named("test_post_phi")
test_post_phi.ObservedValue =System.Array[Vector]([post_phi[k].GetMean() for k in range(K)])

and break symmetry by initialising theta marginals of the test data,i get my result.
- Marked as answer by Microsoft Research Friday, June 3, 2011 5:27 PM
Friday, June 3, 2011 5:27 PM

All replies

0

Sign in to vote

minka replied on 11-21-2009 10:53 AM

The problem is that the first part of the code defines test_post_phi as an ordinary variable (not an array). The last part of the code tries to use test_post_phi as an array. You need to define test_post_phi using Variable.Array, then use SetTo to set its distribution. The SetTo method is described on this page: http://research.microsoft.com/en-us/um/cambridge/projects/infernet/docs/Working%20with%20arrays%20and%20ranges.aspx

Friday, June 3, 2011 5:27 PM
0

Sign in to vote
shakey replied on 11-21-2009 11:52 PM

Thanks for the reply.

I tried as you said using the SetTo method ,but the VariableArray[Vector].SetTo in IronPython does't seems to accept params other than the VariableArray[Vector]

type,I tried many times but failed. However, I find another way around:

post_phi=engine.Infer(phi)
test_post_phi = Variable.Array[Vector](TopicsNum).Named("test_post_phi")
test_post_phi.ObservedValue =System.Array[Vector]([post_phi[k].GetMean() for k in range(K)])

and break symmetry by initialising theta marginals of the test data,i get my result.
- Marked as answer by Microsoft Research Friday, June 3, 2011 5:27 PM
Friday, June 3, 2011 5:27 PM