[UAI] on the inference with Mixture of Graphical Models

From: Gang LI (gangli@bigfoot.com)
Date: Thu Mar 23 2000 - 14:20:14 PST

  • Next message: Philosophy and AI: "[UAI] Bayesianism 2000"

    Please,

    I would like to know if could anyone indicate references about the
    inference with Mixture Graphical Models, like Mixture of Trees (by
    Marina Meila-Predoviciu).

    I suspect that the assertion ("the inference result of a mixture of
    trees is a mixture of the inference of the component trees", page-44,
    in Marina's dissertation) may be incorrect, for inference involves not
    only a marginalization, but also a conditioning operation.

    If this property doesn't hold in the inference with mixture of DAG, is
    it practical to apply this kind of model in real-world tasks? The
    inference speed may be unsatisfactory.

    Best Regards,
    Gang LI
    2000-03-23
    _______________________________________________________________________
    Email: gangli@bigfoot.com
    Department of Computer Science Institute of Software
    Shanghai University Chinese Academy of Sciences
    149#, Yan Chang Road 4#, 4th Road South Zhongguancun
    Shanghai 200072, P.R.China Beijing 100080, P.R.China

    This archive was generated by hypermail 2b29 : Thu Mar 23 2000 - 14:22:44 PST