"""
NLL Laplace Scorer
------------------
Laplace distribution negative log-likelihood Scorer.
The anomaly score is the negative log likelihood of the actual series values
under a Laplace distribution estimated from the stochastic prediction.
"""
import numpy as np
from scipy.stats import laplace
from darts.ad.scorers.scorers import NLLScorer
[docs]class LaplaceNLLScorer(NLLScorer):
def __init__(self, window: int = 1) -> None:
"""NLL Laplace Scorer
Parameters
----------
window
Integer value indicating the size of the window W used by the scorer to transform the series into an
anomaly score. A scorer will slice the given series into subsequences of size W and returns a value
indicating how anomalous these subset of W values are. A post-processing step will convert this anomaly
score into a point-wise anomaly score (see definition of `window_transform`). The window size should be
commensurate to the expected durations of the anomalies one is looking for.
"""
super().__init__(window=window)
def __str__(self):
return "LaplaceNLLScorer"
def _score_core_nllikelihood(
self, vals: np.ndarray, pred_vals: np.ndarray
) -> np.ndarray:
# ML estimate for the Laplace loc
loc = np.median(pred_vals, axis=1)
# ML estimate for the Laplace scale
# see: https://github.com/scipy/scipy/blob/de80faf9d3480b9dbb9b888568b64499e0e70c19/scipy
# /stats/_continuous_distns.py#L4846
scale = np.sum(np.abs(pred_vals.T - loc), axis=0).T / pred_vals.shape[1]
return -laplace.logpdf(vals, loc=loc, scale=scale)