Alpha quickstart

This example uses the immutable v0.1.0a1 API.

import featuregraph as fg

bidmc = fg.datasets.bidmc(subject=1)

oscillation = fg.oscillation.Oscillation(
    signals="respiration",
    group="subject",
    smooth_signal=False,
)

features = oscillation.fit_transform(bidmc)
objects = oscillation.summarize(features, signal="respiration")

The construction retains both the object table and the sample-level evidence:

objects.table
objects.features
objects.construction

Once the oscillations are explicit, query their properties without detecting their boundaries again:

long_oscillations = (
    objects.query()
    .where(duration__ge=100)
    .select(
        "oscillation_id",
        "start_index",
        "end_index",
        "duration",
        "amplitude",
    )
    .collect()
)

Pandas can filter a table. FeatureGraph constructs the behavioral table that makes the filtering meaningful.

For the full example, see the alpha demonstration notebook.