Abstract
We propose a unifying family of quadratic cost functions to be used in Peer-to-Peer ratings. We show that our approach is general since it captures many of the existing algorithms in the fields of visual layout, collaborative filtering and Peer-to-Peer rating, among them Koren spectral layout algorithm, Katz method, Spatial ranking, Personalized PageRank and Information Centrality. Besides of the theoretical interest in finding common basis of algorithms that where not linked before, we allow a single efficient implementation for computing those various rating methods. We introduce a distributed solver based on the Gaussian Belief Propagation algorithm which is able to efficiently and distributively compute a solution to any single cost function drawn from our family of quadratic cost functions. By implementing our algorithm once, and choosing the computed cost function dynamically on the run we allow a high flexibility in the selection of the rating method deployed in the Peer-to-Peer network. Using simulations over real social network topologies obtained from various sources, including the MSN Messenger social network, we demonstrate the applicability of our approach. We report simulation results using networks of millions of nodes.
Similar content being viewed by others
Notes
An extension to non-symmetric matrices is discussed in Bickson et al. [23]. For simplicity of arguments, we handle the symmetric case in this paper.
References
Moallemi CC, Van Roy B (2006) Consensus propagation. IEEE Trans Inf Theory 52(11):4753–4766
Johnson J, Malioutov D, Willsky A (2005) Walk-sum interpretation and analysis of Gaussian Belief Propagation. In: NIPS 05’, Vancouver, 5–10 December 2005
Grinstead CM, Laurie Snell J (1997) Introduction to probability. The CHANCE Project. Chapter 11 - Markov chains. http://www.dartmouth.edu/ chance/teaching_aids/books_articles/probability_book/pdf.html
Brin S, Page L (1998) The anatomy of a large-scale hypertextual Web search engine. In: Proceedings of the seventh international conference on World Wide Web, vol 7. Elsevier, Amsterdam, pp 107–117
Weiss Y, Freeman WT (1999) Correctness of belief propagation in Gaussian graphical models of arbitrary topology. In: NIPS-12, Denver, 30 November–2 December 1999
Malioutov DM, Johnson JK, Willsky AS (2006) Walk-sums and belief propagation in Gaussian graphical models. J Mach Learn Res 7:2031–2064 (October)
Crammer K, Singer Y (2001) Pranking with ranking. In: Proceedings of the conference on neural information processing systems (NIPS)
Doyle PG, Snell JL (1984) Random walks and electrical networks. The mathematical association of America. http://arxiv.org/abs/math.PR/0001057
Brandes U, Fleisch D (2005) Centrality measures based on current flow. In: STACS 2005, LNCS 3404, Stuttgart, 24–26 February 2005, pp 533–544
Benczur A, Csalogany K, Sarlos T (2005) On the feasibility of low-rank approximation for personalized PageRank. In: Poster proceedings of the 14th international World Wide Web conference (WWW), Chiba, 10–14 May 2005, pp 972–973
Bickson D, Dolev D, Shental O, Siegel PH, Wolf JK (2007) Linear detection via belief propagation. In: The 45th annual allerton conference on communication, control, and computing, Allerton House, Illinois, September 2007
Koren Y (2003) On spectral graph drawing. In: Proceedings of the 9th international computing and combinatorics conference (COCOON’03). Lecture notes in computer science, vol 2697. Springer, Berlin Heidelberg New York, pp 496–508
Bickson D, Malkhi D, Zhou L (2007) Peer to peer rating. In: The 7th IEEE Peer-to-Peer Computing, Galway, Ireland, September 2007
Katz L (1953) A new status index derived from sociometric analysis. Psychometrika 18:39–43
Hall KM (1970) An r-dimensional quadratic placement algorithm. Manage Sci 17:219–229
DellAmico M (2007) Mapping small worlds. In: The 7th IEEE Peer-to-Peer computing, Galway, Ireland, September 2007
Johnson JK, Malioutov D, Willsky AS (2007) Lagrangian relaxation for MAP estimation in graphical models. In: The 45th annual Allerton conference on communication, control and computing, Monticello, September 2007.
Bell R, Koren Y (2007) Scalable collaborative filtering with jointly derived neighborhood interpolation weights. In: IEEE international conference on data mining (ICDM’07). IEEE, Piscataway
Elidan G, McGraw I, Koller D (2006) Residual belief propagation: informed scheduling for asynchronous message passing. In: Proceedings of the twenty-second conference on uncertainty in AI (UAI), Cambridge, 13 July 2006
DIMES (2006) The DIMES project. http://www.netdimes.org
Web research collections. http://ir.dcs.gla.ac.uk/test_collections
Pajek - A program for large network analysis. http://vlado.fmf.unilj.si/pub/networks/pajek/
Bickson D, Shental O, Dolev D, Siegel PH, Wolf JK (2008) Gaussian belief propagation based multiuser detection. In: IEEE Int. Symp. Inform. Theory (ISIT), July 2008, Toronto, Canada (to appear)
Acknowledgements
We would like to thank Yaacov Fernandess for his vision and support of the Nocturnal project, and Yair Weiss for being a great research mentor in the field of Gaussian Belief Propagation. We further like to thank Ciamak C. Moallemi, Benjamin Van-Roy, Erik Aurell, Danny Dolev and Jason K. Johnson for interesting discussions regarding the GaBP algorithm. We thank Elad Yom-Tov for supplying the Blogs crawl data and Udi Weinsberg for his support in using the DIMES datasets. We also thank Matteo Dell’Amico for pointing out the relation between visual layouts and node ratings. We thank Ori Shental for clearly explaining the GaBP algorithm in Section 3. Finally, we thank the anonymous reviewers for their helpful comments.
Author information
Authors and Affiliations
Corresponding author
Additional information
Danny Bickson was partially supported by Microsoft Research Internship and by EVERGROW IP 1935 of the EU Sixth Framework.
Appendix
Appendix
Proof of Theorem 1
It is shown in Hall [15] that the optimal solution to the cost function is x = L − 1 where L is the graph Laplacian. Substitute β = 1, w ii = 1, w ij = 1, y i = 1 in the cost function (1) :
The same cost function in linear algebra form:
Now we calculate the derivative and compare to zero and get
□
Proof of Theorem 2
Using the notations of Dell’Amico [16] the cost function of Koren’s spectral layout algorithm is:
s.t
we create a weighted-sum objective:
Taking β = 1, γ = 1/2 we get:
Reverting to our cost function formulation we get:
In other words, taking w ii = deg(i), y i = 1, β = 1 and we get Koren’s formulation.□
It is interesting to note, that the optimal solution according to Koren’s work is \(x_i = \sum_{j \in N(i)}\frac{w_{ij}x_j}{deg(i)}\) which is equivalent to the thin plate model image processing and PDEs [17].
Proof of Theorem 5
We have shown that the fundamental matrix is equal to (I − R) − 1. Assume that the edge weights are probabilities of Markov-chain transitions (which means that each row sums into one), substitute β = α, w ii = 1, y i = 1 in the cost function (1) :
The same cost function in linear algebra form:
Now we calculate the derivative and compare to zero and get
□
Proof of Theorem 6
The proof is very similar to the Spatial Ranking proof. There are two differences: the first is that the prior distribution x is set in y to weight the output towards the prior. Second, in the in the Personalized PageRank algorithm the result is multiplied by the constant (1 − α), which we omit in our cost function. This computation can be done locally at each node after the algorithm terminates, since α is a known fixed system parameter.□
Rights and permissions
About this article
Cite this article
Bickson, D., Malkhi, D. A unifying framework of rating users and data items in peer-to-peer and social networks. Peer-to-Peer Netw. Appl. 1, 93–103 (2008). https://doi.org/10.1007/s12083-008-0008-4
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s12083-008-0008-4