affinity propagation算法Matlab代码实现.docVIP

%APCLUSTER Affinity Propagation Clustering (Frey/Dueck, Science 2007) % [idx,netsim,dpsim,expref]=APCLUSTER(s,p) clusters data, using a set % of real-valued pairwise data point similarities as input. Clusters % are each represented by a cluster center data point (the exemplar). % The method is iterative and searches for clusters so as to maximize % an objective function, called net similarity. % % For N data points, there are potentially N^2-N pairwise similarities; % this can be input as an N-by-N matrix s, where s(i,k) is the % similarity of point i to point k (s(i,k) needn抰 equal s(k,i)). In % fact, only a smaller number of relevant similarities are needed; if % only M similarity values are known (M N^2-N) they can be input as % an M-by-3 matrix with each row being an (i,j,s(i,j)) triple. % % APCLUSTER automatically determines the number of clusters based on % the input preference p, a real-valued N-vector. p(i) indicates the % preference that data point i be chosen as an exemplar. Often a good % choice is to set all preferences to median(s); the number of clusters % identified can be adjusted by changing this value accordingly. If p % is a scalar, APCLUSTER assumes all preferences are that shared value. % % The clustering solution is returned in idx. idx(j) is the index of % the exemplar for data point j; idx(j)==j indicates data point j % is itself an exemplar. The sum of the similarities of the data points to % their exemplars is returned as dpsim, the sum of the preferences of % the identified exemplars is returned in expref and the net similarity % objective function returned is their sum, i.e. netsim=dpsim+expref. % % [ ... ]=apcluster(s,p,NAME,VALUE,...) allows you to specify % optional parameter name/value pairs as follows: % % maxits maximum number of iterations (default: 1000) % convits if the estimated exemplars stay fixed for convits % iterations, APCLUSTER terminates early (default: 100) % dampfac