Expectation-Maximization Binary Clustering for Behavioural Annotation
Fig 5
Performance comparison among EMbC, EMC, and HMM.
Average performance of EMbC, EMC and HMM on 100 synthetic trajectories drawn from a GMM (4 components) using a Markov-chain-based sampling scheme (top panel) and a prior-based sampling scheme (bottom panel), for different trajectory sizes (n = {50, 100, 200, 400, 800, 1600}) and definition of the binary regions (γ = {0.01, 0.05, 0.10}). Values of performance are given in terms of F-measure.