Reviews and Comments on Paper 226
Paper information
| Paper #226: Nobuyuki Taga and Shigeru Mase. Applications of Gibbs Measure Theory to Loopy Belief Propagation Algorithm |
| Abstract: In this paper, we pursue application of Gibbs measure theory to LBP in two ways. First,
we show this theory can be applied directly to LBP for factor graphs,
where one can use higher-order potentials. Consequently, we show
beliefs are just marginal probabilities for a certain Gibbs measure
on a computation tree. We also give a convergence criterion using this tree.
Second, to see the usefulness of this approach,
we apply a well-known general condition and a special one,
which are developed in Gibbs measure theory, to LBP.
We compare these two criteria and another criterion derived by the best
present result.
Consequently, we show that the special condition is better than the others
and also show the general condition is better than the best present result
when the influence of one-body potentials is sufficiently large.
These results surely encourage the use of Gibbs measure theory in this area. (file) |
Summary of received reviews and comments
Reviews superseded by other reviews are shown in the grey color in the table.
| confidence | score | ||||
| Review 1 | 3 | 3 | |||
| Review 2 | 3 | 2 | |||
| Review 3 | 2 | 2 | |||
Reviews and Comments
Review 1
| PC member: | |
| Overall rating: | 3 (strong accept) |
| Confidence: | 3 |
| Relevance: Is this paper relevant for this conference? | 3 (strong accept) |
| Soundness: Is this paper technically sound and complete? | 3 (strong accept) |
| Are the claims sufficiently supported by experimental/theoretical results? | 3 (strong accept) |
| Significance: Are the results/ideas interesting for other AI researchers? | 3 (strong accept) |
| Originality: Are the results or ideas novel and previously unknown? | 2 (accept (I will argue for this paper)) |
| Readability: Is the paper well-organized and easy to understand? | 3 (strong accept) |
| Language: Is the paper written in correct English and style? | 3 (strong accept) |
| Format: Is the paper correctly and consistently formatted? | 3 (strong accept) |
| Review: | CONTRIBUTION OF THE PAPER: Gibbs Measure Theory is used to handle Belief Propagation under the existence of loops in belief networks. POSITIVE ASPECTS: The authors introduce the notion of factor graph and computation trees for belief propagation, and they prove probabilistic properties realized as Gibbs measures on the computation trees, as well as some convergence criteria. NEGATIVE ASPECTS: It is an interesting and very well written paper. There lacks a pronoun at the right beginning of section 2.1 and the use of the same symbol "Z" for several normalizing constants is a little bit confusing in spite of authors preventions. CHANGES TO IMPROVE THE PAPER: FURTHER COMMENTS: ITEMS BELOW ARE JUSTIFICATION OF THE SCORES IF NEGATIVE: (1) IS THIS PAPER RELEVANT FOR THIS CONFERENCE? Yes (2) IS THIS PAPER TECHNICALLY SOUND AND COMPLETE? Yes (3) ARE THE CLAIMS SUFFICIENTLY SUPPORTED BY EXPERIMENTAL OR THEORETICAL RESULTS? Yes (4) ARE THE RESULTS/IDEAS INTERESTING FOR OTHER AI RESEARCHERS? Yes (5) ARE THE RESULTS OR IDEAS NOVEL AND PREVIOUSLY UNKNOWN? Yes (6) IS THE PAPER WELL-ORGANIZED AND EASY TO UNDERSTAND? Yes (7) IS THE PAPER WRITTEN IN CORRECT ENGLISH AND STYLE? Yes (8) IS THE PAPER CORRECTLY AND CONSISTENTLY FORMATTED? Yes |
| PC only: | |
| Time: | Jun 30, 20:01 |
Review 2
| PC member: | |
| Reviewer: | |
| Overall rating: | 2 (accept: I will argue for this paper) |
| Confidence: | 3 |
| Relevance: Is this paper relevant for this conference? | 2 (accept (I will argue for this paper)) |
| Soundness: Is this paper technically sound and complete? | 2 (accept (I will argue for this paper)) |
| Are the claims sufficiently supported by experimental/theoretical results? | 2 (accept (I will argue for this paper)) |
| Significance: Are the results/ideas interesting for other AI researchers? | 2 (accept (I will argue for this paper)) |
| Originality: Are the results or ideas novel and previously unknown? | 1 (weak accept (vote accept but don't mind rejecting)) |
| Readability: Is the paper well-organized and easy to understand? | -1 (weak reject (vote reject but don't mind accepting)) |
| Language: Is the paper written in correct English and style? | -1 (weak reject (vote reject but don't mind accepting)) |
| Format: Is the paper correctly and consistently formatted? | 1 (weak accept (vote accept but don't mind rejecting)) |
| Review: | CONTRIBUTION OF THE PAPER: The authors present two applications of Gibbs measure theory to LBP algorithms POSITIVE ASPECTS: The theory can be applied to probability functions with general potentials through factor graphs. The authors claim that an application of this theory gives a better results that the best current results in special cases NEGATIVE ASPECTS: How valid is the claim that the results are better just for special cases rather than for more general cases? CHANGES TO IMPROVE THE PAPER: The English writing should be revised and improved FURTHER COMMENTS: The problem is interesting, the authors are well documented and present a valid algorithm for special cases of LBP. ITEMS BELOW ARE JUSTIFICATION OF THE SCORES IF NEGATIVE: (1) IS THIS PAPER RELEVANT FOR THIS CONFERENCE? yes (2) IS THIS PAPER TECHNICALLY SOUND AND COMPLETE? seems to be (3) ARE THE CLAIMS SUFFICIENTLY SUPPORTED BY EXPERIMENTAL OR THEORETICAL RESULTS? seems to be (4) ARE THE RESULTS/IDEAS INTERESTING FOR OTHER AI RESEARCHERS? think so (5) ARE THE RESULTS OR IDEAS NOVEL AND PREVIOUSLY UNKNOWN? not quite (6) IS THE PAPER WELL-ORGANIZED AND EASY TO UNDERSTAND? needs improvements (7) IS THE PAPER WRITTEN IN CORRECT ENGLISH AND STYLE? should be improved (8) IS THE PAPER CORRECTLY AND CONSISTENTLY FORMATTED? is ok |
| PC only: | Seems like a good paper |
| Time: | Jul 13, 19:25 |
Review 3
| PC member: | |
| Reviewer: | |
| Overall rating: | 2 (accept: I will argue for this paper) |
| Confidence: | 2 |
| Relevance: Is this paper relevant for this conference? | 2 (accept (I will argue for this paper)) |
| Soundness: Is this paper technically sound and complete? | 2 (accept (I will argue for this paper)) |
| Are the claims sufficiently supported by experimental/theoretical results? | 2 (accept (I will argue for this paper)) |
| Significance: Are the results/ideas interesting for other AI researchers? | 1 (weak accept (vote accept but don't mind rejecting)) |
| Originality: Are the results or ideas novel and previously unknown? | 2 (accept (I will argue for this paper)) |
| Readability: Is the paper well-organized and easy to understand? | 3 (strong accept) |
| Language: Is the paper written in correct English and style? | 3 (strong accept) |
| Format: Is the paper correctly and consistently formatted? | 3 (strong accept) |
| Review: | CONTRIBUTION OF THE PAPER: This paper shows that Gibbs measure theory can be applied to loopy belief propagation algorithms (LBP) defined for factor graphs. A convergence criterion is given for computation trees defined from factor graphs. POSITIVE ASPECTS: - Definition of computation trees for factor graphs allowing to analyze LBP for higher order potentials. - Results on the application of the Gibbs measure to computation trees with higher order potentials. NEGATIVE ASPECTS: - The description of the procedure to check the Simon's condition is obscure. For instance, it is not clear how to fix factor A at STEP 2 of proposition 3. Which is the influence of the order selected to fix the factors? - The comparison with other convergence criteria is done for only one case with pair potentials. It is not clear how the method behave for actual higher order potentials. Therefore, the significance of the paper is not clear. CHANGES TO IMPROVE THE PAPER: - See comments above - Some more discussion on the use of Gibbs measure theory for the study of LBP would make the signicancy of your paper clearer. For example, authors state that "Nevertheless, to use this (Gibbs measure) theory seems not to be so popular". Why? Could you provide information to the reader? Are there alternative theoretical frameworks for the study of LBP? How do they compare to the approach chosen by the authors? FURTHER COMMENTS: ITEMS BELOW ARE JUSTIFICATION OF THE SCORES IF NEGATIVE: (1) IS THIS PAPER RELEVANT FOR THIS CONFERENCE? (2) IS THIS PAPER TECHNICALLY SOUND AND COMPLETE? (3) ARE THE CLAIMS SUFFICIENTLY SUPPORTED BY EXPERIMENTAL OR THEORETICAL RESULTS? (4) ARE THE RESULTS/IDEAS INTERESTING FOR OTHER AI RESEARCHERS? (5) ARE THE RESULTS OR IDEAS NOVEL AND PREVIOUSLY UNKNOWN? (6) IS THE PAPER WELL-ORGANIZED AND EASY TO UNDERSTAND? (7) IS THE PAPER WRITTEN IN CORRECT ENGLISH AND STYLE? (8) IS THE PAPER CORRECTLY AND CONSISTENTLY FORMATTED? |
| PC only: | |
| Time: | Jul 14, 09:46 |