Negative log likelihood loss of HCRP models fitted to 10.000 ASRT data points.

(a) Negative log likelihood as a function of the maximum number of previous events considered. (b) Negative log likelihood as a function of the prior importance of two previous events, i.e. trigrams (b). In (b), lower values of α2 imply higher prior importance. The vertical dashed line in (a) marks the n that was used for fitting the human data in the Manuscript. (TIFF)