Skip to contents

Soft Dynamic Time Warping Distance

Source: R/metric.R
metric.softDTW.Rd

Computes the Soft-DTW distance matrix between functional data objects. Soft-DTW is a differentiable relaxation of DTW that uses a smoothing parameter gamma to control the softness of the minimum operation.

Usage

metric.softDTW(fdataobj, fdataref = NULL, gamma = 1, divergence = FALSE, ...)

Arguments

fdataobj

An object of class 'fdata'.

fdataref

An object of class 'fdata'. If NULL, computes self-distances.

gamma

Smoothing parameter (> 0). Smaller values approximate hard DTW. Default is 1.0.

divergence

Logical. If TRUE, computes the Soft-DTW divergence (non-negative, zero for identical series) instead of raw Soft-DTW distance. Default is FALSE.

...

Additional arguments (ignored).

Value

A distance matrix.

References

Cuturi, M. and Blondel, M. (2017). Soft-DTW: a Differentiable Loss Function for Time-Series. Proceedings of the 34th International Conference on Machine Learning (ICML).

Examples

fd <- fdata(matrix(rnorm(100), 10, 10))
D <- metric.softDTW(fd, gamma = 1.0)
D_div <- metric.softDTW(fd, gamma = 1.0, divergence = TRUE)