extern double movie_vote(Prediction. Config *pcfg, unsigned long int M, PTree & support. M, unsigned long int U, PTree & support. U) { auto double MU=Users. get_rating(U, M)-2, VOTE=DEFAULT_VOTE, VOTE_sum=0, VOTE_cnt=0, Nb, Mb, ds. Sq, UCor=1, support. Usize=support. U. get_count(), support. Msize=support. M. get_count(); struct pruning *internal_prune; struct external_prune *external_prune; auto PTree sup. M=support. M, sup. U=support. U; sup. M. clearbit(U); sup. U. clearbit(M); movie-vote. C ARM code 1 auto PTree. Set & U_ptree_set= Users. get_ptreeset(), & M_ptree_set=Movies. get_ptreeset(); sup. U. clearbit(M); sup. M. clearbit(U); auto PTree sup. U_1=sup. U&(~U_ptree_set[(U*3)+0])&( U_ptree_set[(U*3)+1])&( U_ptree_set[(U*3)+2]), sup. U_2=sup. U&( U_ptree_set[(U*3)+0])&(~U_ptree_set[(U*3)+1])&(~U_ptree_set[(U*3)+2]), sup. U_3=sup. U&( U_ptree_set[(U*3)+0])&(~U_ptree_set[(U*3)+1])&( U_ptree_set[(U*3)+2]), sup. U_4=sup. U&( U_ptree_set[(U*3)+0])&( U_ptree_set[(U*3)+1])&(~U_ptree_set[(U*3)+2]), sup. U_5=sup. U&( U_ptree_set[(U*3)+0])&( U_ptree_set[(U*3)+1])&( U_ptree_set[(U*3)+2]), sup. M_1=sup. M&(~M_ptree_set[(M*3)+0])&( M_ptree_set[(M*3)+1])&( M_ptree_set[(M*3)+2]), sup. M_2=sup. M&( M_ptree_set[(M*3)+0])&(~M_ptree_set[(M*3)+1])&(~M_ptree_set[(M*3)+2]), sup. M_3=sup. M&( M_ptree_set[(M*3)+0])&(~M_ptree_set[(M*3)+1])&( M_ptree_set[(M*3)+2]), sup. M_4=sup. M&( M_ptree_set[(M*3)+0])&( M_ptree_set[(M*3)+1])&(~M_ptree_set[(M*3)+2]), sup. M_5=sup. M&( M_ptree_set[(M*3)+0])&( M_ptree_set[(M*3)+1])&( M_ptree_set[(M*3)+2]), sou. M, sou. U, som. M, sp. U; auto double thr 1, expnt 1, thr 2, expnt 2, s, S, ss, sn, s. M, s. U, c, C, wt, XBal. VT, wt_const=16; //SAMPLE-stat d. MNsds pruning parms hijacked for ARM parm. /* External pruning: Prune Users sup. M */ internal_prune = pcfg->get_internal_prune(movie_d. MNsds); external_prune=pcfg->get_movie_Prune_Users_in_Sup. M(); if (external_prune->enabled) { thr 1=internal_prune->threshold; if(sup. M. get_count()>external_prune->params. Ct) expnt 1=internal_prune->exponent; do_pruning(external_prune, M, U, sup. M, sup. U); internal_prune=pcfg->get_internal_prune(movie_Nsds_Msds); sup. M. clearbit(U); sup. U. clearbit(M); if((sup. M. get_count()<1)||(sup. U. get_count()<1)) return thr 2=internal_prune->threshold; VOTE; } expnt 2=internal_prune->exponent; /* External pruning: Prune Movies sup. U */ external_prune=pcfg->get_movie_Prune_Movies_in_Sup. U(); if (external_prune->enabled) { if(sup. U. get_count()>external_prune->params. Ct) do_pruning(external_prune, M, U, sup. M, sup. U); sup. M. clearbit(U); sup. U. clearbit(M); if((sup. M. get_count()<1)||(sup. U. get_count()<1)) return VOTE;
KDD-Cup 2011: Recommending Music Items based on the Yahoo! Music Dataset We challenge participants to identify user tastes in music by analyzing real ratings of Yahoo! Music anonymized users. The dataset represents a snapshot of the community's preferences for various musical items. A distinctive feature of this dataset is that user ratings are not given to a single type of entities as is usually done but to four different types, namely tracks, albums, artists, and genres, tied together within a hierarchy. Thus, any given track is associated with its album, performing artist and associated genres. Similarly, any given album is associated with an artist and genre(s). The competition offers two tracks, differing by the dataset size and the accuracy metric. The first track employs a dataset containing over 260 M ratings. For this dataset the task is to predict test set ratings as accurately as possible. A second track concentrates on a smaller training set with about 62 M train ratings. Here, the goal is to separate items rated highly by the users from items never rated by the users. The main technique we expect participants to use is collaborative filtering. We believe that more successful attempts will require novel techniques and approaches, as this challenge dataset pushes the limits of current recommender systems in several dimensions: Structure: The Yahoo! Music dataset comprises typed items, which are all linked together within a defined hierarchy. Time: The Yahoo! Music dataset reports rating times, which allows performing session analysis of user activities and to infer exact chronological order of ratings. Scale: The Yahoo! Music dataset is very large with as many as 625 K items, much larger than any similar public dataset, where usually only the number of users is large. It also includes significantly more ratings than the currently largest one - the Netflix Prize dataset. Important Dates (tentative) March 1, 2011: Registration opens March 15, 2011: Competition begins June 30, 2011: Competition ends August 21, 2011: Workshop and winners presentation Workshop Following KDD-Cup’s tradition we will hold a workshop during KDD’s opening day. During the workshop we will invite the competition leaders to present their techniques, together with several external speakers. Workshop details will be published later. Organizing Committee Gideon Dror : Yahoo! Labs, Israel Yehuda Koren: Yahoo! Labs, Israel Markus Weimer, Yahoo! Labs, US
Скачать презентацию extern double movie_vote Prediction Config pcfg unsigned long int
685c43ee157aad0fedfa71e80088d124.ppt