Regressing / smoothing input time-series based on anomalies

[arun-nemani]

Is there a way to objectively regress / normalize discrete points in the original time-series (ts) based on the anomalies time series (spikes), which are essentially "weights". I basically want to use the anomaly detector as a smoothing mask. Does this exist currently?

detector = anomaly_detector.AnomalyDetector(ts)
spikes = detector.get_all_scores().values