s6.utils.filters

Generic temporal smoothing filters.

This module provides a lightweight sliding-window smoother that accepts lists/tuples/arrays of arbitrary length and returns a moving-average value. An optional acceptance function allows rejecting outlier measurements before they are added to the window, which is useful for stabilizing detectors.

class s6.utils.filters.TupleSmoother(window: int = 3, accept_fn: ~typing.Callable[[~numpy.ndarray, ~numpy.ndarray], bool] | None = None, dtype: ~numpy.dtype = <class 'numpy.float64'>)

Bases: object

Sliding-window smoother for numeric tuples/lists/arrays.

  • Accepts 1D numeric sequences (list/tuple/ndarray) of any length.

  • Maintains a fixed-size history of accepted measurements.

  • Returns the moving average over the history as the smoothed value.

  • Supports an optional accept_fn(new, prev) -> bool predicate to reject outliers before they are incorporated into the history.

reset() → None
property window: int
property has_value: bool
property history: List[ndarray]

Return a list copy of internal history arrays (oldest..newest).

property value: ndarray | None

Return the current smoothed value (moving average) or None if empty.

update(value: Sequence[float]) → Tuple[ndarray, bool]

Push a new measurement and return (smoothed_value, accepted).

The measurement is only appended to history if accept_fn allows it (or if there is no prior history). The smoothed value is the moving average over the current history window.