At present about personalized PageRank, other common methods still have a model to change PageRank(modular PageRank) and BlockRank to wait. These methods are on specific computational means, main characteristic body is opposite from the angle of efficiency now algorithm undertook be optimizinged necessarily.

About accelerating the previous research of PageRank algorithm content basically uses technology of texture of picture of few and far between sex, for instance the Arasu point of view that offer, they not just pure use last iteration produces a value to calculate circularly epicycle loop value, the value that also uses epicycle to had arisen circularly will quicken the computation that epicycle circulates. Offerred the bowknot structure of Web network even, use its in the active consideration that PageRank is worth. However these methods are not had very big practical, main reason is asking to have sort to Web network matrix at algorithm, this operation needs to search first principle to undertake the network alls over all previous according to deepness, this is operation of huge of a kind of cost apparently. The closest Kamvar also put forward a few algorithm, use successive intermediate loop will conclude the estimation value with true better PageRank, but the part that still is put in the inadequacy that suffers effect of PageRank algorithm initiative parameter.

Basically pay close attention to the attribute that pursues at research to the analysis of texture of Web network plan at present, be like node distributing, of the circumstance that the webpage links and texture of Web webpage picture build a model to wait. However, will accelerate analysis exceeding cable length to these study and emphasizing how using these property effectively.

Many scholars put forward a few improve a practice, wait like Raghavan and Garcia-Molina use leader name or the practice that the Web structure with implicit URL will come pursue more successful on behalf of Web also has a lot of, the personalized webpage weight that if Jeh and Widom carry the power of finite modification webpage,the value is conveyed, worth of this importance power can mirror the initiative interest webpage that the user appoints. Because need to all over all previous repeatedly to the computation of personalized view,whole Web pursues the webpage in the structure, during this is moving only, ability comes true, be calculated beforehand so and store all personalized view are not realistic. The preference vector that they use new plan to build an expressive individuation view by result and technical compose (Partial Vector) , it is OK share in the personalized view of different user, be calculated about it at the same time and of memory expenditure and view amount more or less to present a reasonable scale. In computation, still can use increase by degrees type is calculated, this makes preference vector is used to go during inquiry it is feasible that compose builds personalized view. This preference vector is personalized PageRank vector namely (Personalized PageRank Vector, PPV) , common ground says, PPV is the personalized view of kind of Web webpage. Will undertake sort express the preference of the user effectively to webpage result according to this PPV.

Look simply, the length of every PPV is cuss, namely the webpage amount of Web. But because from fixed angle circulates,computational PPV need alls over graph of webpage of all previous Web for many times, this is impossible that as a kind online noise uses the means of an inquiry apparently. From the point of another angle, the total amount of all PPV vector can achieve 2n(n to be webpage gross) , this is apparent too tremendous and cannot realize memory leaving a line. So, must restrict the webpage that appears in P gather the subclass that is H of Hub webpage gather. H gather includes the webpage that a few users are interested in most normally. In practice, h gather can be the webpage rally that has taller PageRank value (important webpage) , the webpage in classifying catalog artificially (the important webpage that is like Yahoo and Open Directory) , specific company or program. H gather can regard the foundation of computational individuation as. This kind is based on the computational means of PPV, the means of unlike tradition, can become good proportion to shrink with H gather size put a concern, and this kind of technology also can be in bigger PPV obtains similar result on gather, a few more contented to individuation of aggregate of aleatoric preference webpage computation asks.

Divide this beyond, still a few undertake improved algorithm on computational effect.