Short movement patterns reveal animal identity and learning

Researchers found that movement trajectories as short as one second can act as a fingerprint for individual rats.

Short-term movement can reveal an animal's identity and its learning progress. These movement patterns act as an informative behavioral fingerprint.

A new framework was built using a self-supervised representation learning framework. This framework uses temporal proximity as supervision. It learns from movement trajectories without manually defining behavioral states or kinematic features.

The researchers used behavior data from freely behaving rats. They found that trajectories as short as one second contain informative signals. These signals include identity and learning-related readouts.

Egocentric pose dynamics are more predictive of identity. Allocentric motion is more predictive of learning-related variation. Both signals are strongest at decision junctions in the alternation cohort.

A representation-conditioned diffusion model was used to inspect movement patterns. This tool helps investigate when and how these signals appear in behavior.

Key facts

  • Movement trajectories as short as one second contain identity and learning signals.
  • Egocentric pose dynamics predict identity better than other measures.
  • Allocentric motion is more predictive of learning-related variation.

All sections

Edition