Is the clustering algorithm eqivalent to the HUGIN junction tree?

The front end.
Post Reply
JohnLee
Posts: 6
Joined: Thu Apr 19, 2018 1:15 pm

Is the clustering algorithm eqivalent to the HUGIN junction tree?

Post by JohnLee » Sat Dec 01, 2018 1:04 am

Is the default inference algorithm "clustering algorithm" the so-called HUGIN junction tree algorithm? or Shafer-shenoy tree algorithm?

marek [BayesFusion]
Site Admin
Posts: 254
Joined: Tue Dec 11, 2007 4:24 pm

Re: Is the clustering algorithm eqivalent to the HUGIN junction tree?

Post by marek [BayesFusion] » Sat Dec 01, 2018 4:12 am

SMILE's clustering algorithm is a variation on Lauritzen-Spiegelhalter's junction tree algorithm. It is a variation and, as you may have noticed, it is very very fast. It is a result of combining good ideas with good engineering.
Cheers,

Marek

JohnLee
Posts: 6
Joined: Thu Apr 19, 2018 1:15 pm

Re: Is the clustering algorithm eqivalent to the HUGIN junction tree?

Post by JohnLee » Sun Dec 02, 2018 5:06 am

Hi Merek,

Could you please give us some relevant materials on the specific inference process? I found the following paper would be useful if anyone is interested.

Lepar V, Shenoy P P. A comparison of Lauritzen-Spiegelhalter, Hugin, and Shenoy-Shafer architectures for computing marginals of probability distributions[J]. arXiv preprint arXiv:1301.7394, 2013.

Thank you very much!

John

marek [BayesFusion]
Site Admin
Posts: 254
Joined: Tue Dec 11, 2007 4:24 pm

Re: Is the clustering algorithm eqivalent to the HUGIN junction tree?

Post by marek [BayesFusion] » Tue Dec 04, 2018 9:24 pm

It took us over twenty years to implement the fastest and most reliable version of the clustering algorithm there is. Do you mean you would like us to disclose the details of this implementation :-)? As I wrote before, it is good theory, most of which has been published in the literature, and great engineering.

Marek

Post Reply