A Sequential Bayesian Alternative to the Classical Parallel Fuzzy Clustering Model

摘要: Unsupervised separation of a group datums particular type, into clusters which are homogenous within problem class-specific context, is classical research still actively visited. Since the 1960s, community has converged class clustering algorithms, utilizes concepts such as fuzzy/probabilistic membership well possibilistic and credibilistic degrees. In spite differences in formalizations approaches to loss assessment different significant majority works literature utilize sum datum-to-cluster distances for all clusters. essence, this double summation basis on additional features outlier rejection robustification built. work, we revisit concept suggest an alternative model function sequentially. We exhibit that notion being emerges mathematical developed paper. Then, provide generic new framework. fact, independent any datum or cluster models robust function. An important aspect work modeling entirely based Bayesian inference framework avoid engineering terms heuristics intuitions. then develop solution strategy functions Alternating Optimization pipeline.