Fig. 7: Dependence of classification accuracy on the used hyper-parameter combination.
From: Sleep as a random walk: a super-statistical analysis of EEG data across sleep stages
Distributions of accuracies over all 68 data sets, with different combinations of hyper-parameters used in the Bayesian likelihood: a Single hyper-parameters. Here, the individual mean accuracies are μSTD = 0.42, μKUR = 0.34, μSKE = 0.44, and the global mean of these three values is μglo = 0.40. b Pairs of hyper-parameters. Here, μSTD,KUR = 0.45, μKUR,SKE = 0.47, μSTF,SKE = 0.48, and the global mean is μglo = 0.47. c All three hyper-parameters. Here the global mean is μglo = 0.51. The global mean is systematically increasing with the number of used hyper-parameters.
