Regressing / smoothing input time-series based on anomalies

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`

Start at anomaly_detector.AnomalyDetector and inspect the get_all_scores().values output to understand what anomaly weights are currently exposed. Determine whether the project already supports using those scores to regress or normalize the original time series; done would require a clearly specified smoothing behavior and corresponding implementation or documentation.

Written by the indexing model from the issue text.