P. Tsaparas. Link Analysis Ranking. PhD thesis, University of Toronto, 2004.  Home/Search   Document Not in Database   Summary   Related Articles

....before any normalization is applied. These two vectors appear to be close. Suppose that we normalize the LAR vectors in the L# norm, and let w# and v# denote the normalized vectors. Then w # (i) v (i) n 1) 2 = #(n) The maximum L 1 distance between any two L# unit vectors is #(n) [52], therefore, these two vectors appear to be far apart. Suppose now that we normalize in the L 1 norm, and let w 1 and v 1 denote the normalized vectors. Then w 1 (i) v 1 (i) 2(n 1) n(n 1) #(1 n) The maximum L 1 distance between any two L 1 unit vectors is #(1) therefore, the two ....

....so that they takes values in [0, 1] We will use the weak rank distance when we want to argue about two LAR vectors being far apart, and the strict rank distance when we want to argue about two LAR vectors being close. The problem of comparing partial rankings is studied independently in [52] and [19] where they discuss about the properties of Kendal tau distance on partial rankings. It is shown that the d r distance measure is a metric. Also, Fagin et al. [19] generalize other distance measures for the case of partial rankings. 4.4 Similarity of LAR algorithms We now turn to the ....

[Article contains additional citation context not shown here]

P. Tsaparas. Link Analysis Ranking. PhD thesis, University of Toronto, 2004.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC