I'm wondering if anybody has followed up on Nair/Tetali's correlation decay tree construction for hypergraphs. I didn't find anything relevant in back-citation on google scholar. Interesting question are:

  1. applying it to exact and approximate inference in Boltzmann machines, and
  2. using it to determine for which hypergraphs it is tractable to count the number of colorings approximately.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.