Fig. 2: Unsupervised clustering in the spatiotemporal feature space.

a Distribution of the signals in the 3D spatiotemporal feature space. Gray dots represent the outliers identified using a distance-based detection method, while remaining data points were labeled in pink. The heatmap shows the distribution density of the remaining data points when projected onto the temporal feature plane. b Dendrogram showing the cluster structure of the remaining phase traces (pink dots in Fig. 2a) in the spatiotemporal feature space, with only the top 100 subclusters displayed. c Clustering the remaining phase traces in the spatiotemporal feature space into three clusters by thresholding the dendrogram along the solid black line. Colored dots denote different groups, and the gray dots denote outliers. d Corresponding representative Type-I and Type-II signals obtained by averaging the individual phase traces within each cluster. The solid lines and color bands denote the mean values and the range of standard deviations. Source data of Fig. 2d are provided as a Source Data file.