Skip to main content
. 2023 Nov 20;27(1):187–195. doi: 10.1038/s41593-023-01490-6

Extended Data Fig. 8. HMM modeling of deep features, keypoints and neural activity.

Extended Data Fig. 8

ac, Test log-likelihood for HMM models trained on deep features (a), keypoints (b) and neural activity clusters (c), as a function of the number of hidden states in the HMM. Curves represent different mice. b, Top: visualization of keypoint data (after z scoring). Bottom: inferred states by the HMM model. c, Reconstructed keypoint data from the inferred HMM states. d, Top: simulated keypoint data from the HMM. Bottom: simulated state dynamics. eg, Same as b–d but for modeling neural activity clusters. Each cluster represents the average activity of 100–300 neurons.