Research reportDetermining shoal membership using affinity propagation

We propose using the affinity propagation (AP) clustering algorithm for detecting multiple disjoint shoals, and we present an extension of AP, denoted by STAP, that can be applied to shoals that fusion and fission across time. STAP incorporates into AP a soft temporal constraint that takes cluster dynamics into account, encouraging partitions obtained at successive time steps to be consistent with each other. We explore how STAP performs under different settings of its parameters (strength of the temporal constraint, preferences, and distance metric) by applying the algorithm to simulated sequences of collective coordinated motion. We study the validity of STAP by comparing its results to partitioning of the same data obtained from human observers in a controlled experiment. We observe that, under specific circumstances, AP yields partitions that agree quite closely with the ones made by human observers. We conclude that using the STAP algorithm with appropriate parameter settings is an appealing approach for detecting shoal fusion-fission dynamics.

► We propose using the affinity propagation clustering for detecting multiple shoals. ► A soft temporal constraint is included in order to detect shoal fusion and fission. ► We explore how affinity propagation performs on agent-based simulated shoals. ► We compare affinity propagation clustering to human clustering of the same data. ► Affinity propagation is an appealing approach for detecting shoal dynamics.