Polars ts

polars_ts

compute_pairwise_distance(input1, input2, method='dtw', **kwargs)

Compute pairwise distance between time series using the specified method.

This is the unified entry point for all distance metrics. It dispatches to the appropriate Rust implementation based on the method parameter.

Parameters:

Name	Type	Description	Default
input1	DataFrame	DataFrame with columns unique_id and y (univariate) or unique_id and multiple value columns (multivariate).	required
input2	DataFrame	DataFrame with the same schema as input1.	required
method	Literal['dtw', 'ddtw', 'wdtw', 'msm', 'erp', 'lcss', 'twe', 'sbd', 'frechet', 'edr', 'dtw_multi', 'msm_multi']	The distance metric to use. One of: Univariate: dtw — Dynamic Time Warping (default). Accepts dtw_method (standard, sakoe_chiba, itakura, fast) and param. ddtw — Derivative DTW. No extra parameters. wdtw — Weighted DTW. Accepts g (weight penalty, default 0.05). msm — Move-Split-Merge. Accepts c (cost, default 1.0). erp — Edit Distance with Real Penalty. Accepts g (gap value, default 0.0). lcss — Longest Common Subsequence. Accepts epsilon (threshold, default 1.0). twe — Time Warp Edit Distance. Accepts nu (stiffness, default 0.001) and lambda_ (edit penalty, default 1.0). Multivariate: dtw_multi — Multivariate DTW. Accepts metric (manhattan or euclidean). msm_multi — Multivariate MSM. Accepts c (cost, default 1.0).	'dtw'
**kwargs	Any	Method-specific parameters (see above).	{}

Returns:

Type	Description
DataFrame	A DataFrame with columns unique_id_1, unique_id_2, and the distance column.

Raises:

Type	Description
ValueError	If an unknown method or unexpected keyword argument is passed.

Examples:

>>> compute_pairwise_distance(df, df, method="dtw")
>>> compute_pairwise_distance(df, df, method="lcss", epsilon=0.5)
>>> compute_pairwise_distance(df, df, method="dtw", dtw_method="sakoe_chiba", param=3.0)

mann_kendall(expr)

Mann-Kendall test for expression.

sens_slope(expr)

Sen's slope estimator (median of pairwise slopes).