Skip to main content
. Author manuscript; available in PMC: 2023 Sep 8.
Published in final edited form as: Nat Comput Sci. 2023 Mar 27;3(3):240–253. doi: 10.1038/s43588-023-00419-0

Fig. 3 |. Alternative embeddings featuring time.

Fig. 3 |

a, Manifold embedding of Sherlock data with PHATE+Time, Smooth PHATE, and T-PHATE, in HV and EA ROIs, colored by time index. b, StudyForrest localizer data embeddings using PHATE+Time, Smooth PHATE, and T-PHATE in EV and HV. Points are colored by the stimulus category presented at a given time point, which are presented in a random order. T-PHATE reveals the best clustering by stimulus category, showing that the manifold successfully denoises data to learn structure in the data also independent of time. c, Support vector classification of stimulus categories presented during the StudyForrest localizer task based on the embedding data. Dots represent individual subjects (n = 14); bars represent the average classification accuracy across subjects;error bars represent the 95% confidence interval of the mean, estimated with 1,000 bootstrap iterations; the dashed line represents chance classification (1/6). See Supplementary Fig. 1b for further benchmark results.