Perturbation analysis for EPHi,<br/>
<br/>
Consider the clutter problem of Tom Minka where the goal is to get posterior of the mean of one Gaussian component. The likelihood is a mixture of two Gaussians, one of which is just noise. The other component is the true data generator. Say I use EP to approximate
this likelihood term with one Gaussian by moment matching. <br/>
<br/>
- What happens to the inferred mean if I perturb the mean/variance of this approximate Gaussian ?
<br/>
- Which one affects the inferred mean more between having an inaccurate mean (in approximate factors) or an inaccurate variance ?<br/>
- In general, is there some work that analyzes how EP behaves as we vary approximate factors ?<br/>
<br/>
Thanks,<br/>
Wittawat<br/>© 2009 Microsoft Corporation. All rights reserved.Fri, 21 Nov 2014 18:34:53 Zd4353e49-081c-4c00-b64b-ee297d6403fc- https://social.microsoft.com/Forums/en-US/d4353e49-081c-4c00-b64b-ee297d6403fc/perturbation-analysis-for-ep?forum=infer.net#d4353e49-081c-4c00-b64b-ee297d6403fchttps://social.microsoft.com/Forums/en-US/d4353e49-081c-4c00-b64b-ee297d6403fc/perturbation-analysis-for-ep?forum=infer.net#d4353e49-081c-4c00-b64b-ee297d6403fcwittawatjhttps://social.microsoft.com:443/profile/wittawatj/?type=forumPerturbation analysis for EPHi,<br/>
<br/>
Consider the clutter problem of Tom Minka where the goal is to get posterior of the mean of one Gaussian component. The likelihood is a mixture of two Gaussians, one of which is just noise. The other component is the true data generator. Say I use EP to approximate
this likelihood term with one Gaussian by moment matching. <br/>
<br/>
- What happens to the inferred mean if I perturb the mean/variance of this approximate Gaussian ?
<br/>
- Which one affects the inferred mean more between having an inaccurate mean (in approximate factors) or an inaccurate variance ?<br/>
- In general, is there some work that analyzes how EP behaves as we vary approximate factors ?<br/>
<br/>
Thanks,<br/>
Wittawat<br/>Fri, 21 Nov 2014 12:57:11 Z2014-11-21T12:57:11Z
- https://social.microsoft.com/Forums/en-US/d4353e49-081c-4c00-b64b-ee297d6403fc/perturbation-analysis-for-ep?forum=infer.net#6f7dd25a-50fe-4eac-98a6-4cd696290554https://social.microsoft.com/Forums/en-US/d4353e49-081c-4c00-b64b-ee297d6403fc/perturbation-analysis-for-ep?forum=infer.net#6f7dd25a-50fe-4eac-98a6-4cd696290554Tom Minkahttps://social.microsoft.com:443/profile/tom%20minka/?type=forumPerturbation analysis for EPThis forum is only for questions about Infer.NET. Your question is better suited to <a href="http://stats.stackexchange.com/">Cross Validated.</a>Fri, 21 Nov 2014 18:34:53 Z2014-11-21T18:34:53Z