Extended Data Fig. 4: Episode and recall topic proportions matrix K-optimization functions.
From: Geometric models reveal behavioural and neural signatures of transforming experiences into memories
We selected the optimal K-value for the episode and each recall topic proportions matrix using the formula described in Methods. This computation resulted in a curve for each matrix, describing the Wasserstein distance between the distributions of within-event and across-event topic vector correlations, as a function of K.
