Expand description
Trajectory distance measures for comparing entity evolutions.
Provides measures that compare sequences of vectors, not individual points. This is a fundamentally different operation from kNN search.
§Available Measures
| Measure | Complexity | Properties |
|---|---|---|
discrete_frechet | O(n×m) | Respects ordering, handles unequal lengths |
| Signature distance | O(K²) | Universal features, very fast (see [signatures]) |
Signature distance is recommended as the default: it’s O(K²) per comparison and captures all order-dependent temporal dynamics. Fréchet is offered as the exact alternative when ordering-sensitive geometry matters.
§References
- Eiter, T. & Mannila, H. (1994). Computing discrete Fréchet distance.
- Toohey, K. & Duckham, M. (2015). Trajectory Similarity Measures. ACM SIGSPATIAL.
Functions§
- discrete_
frechet - Compute the discrete Fréchet distance between two trajectories.
- discrete_
frechet_ temporal - Compute discrete Fréchet distance from timestamped trajectories.
- l2_dist 🔒
- L2 distance between two vectors.