Beat-Type Accuracy Battery - Sensor Visualization

Runs every registered beat type at 100% purity and at 80% N + 20% selected through the full sim → Artella correction → BeatIdentifier classification flow. Goal: reach 100% type-accuracy in the simulated environment.

Sample size

Match tolerance: 150 ms. Each scenario is repeated 1 time(s) with a deterministic seed sequence — runs use different seeds, but the seed set per multiplier level is fixed so the same level always produces the same numbers.

Per beat-type aggregate

Type-accuracy = (ground-truth beats of type X that were detected AND classified X) / (ground-truth beats of type X), aggregated over every scenario where type X appears. Cells <100% are where the classifier currently fails — the gap to close.

GT type	GT count	Detected	Detection rate	Correctly typed	Type accuracy	Most common mis-classifications

Per scenario

One row per simulator configuration. Click a row to expand the per-type breakdown for that scenario.

Scenario	Runs	GT	Detected	Matched	Correct type	Missed	False+	Type accuracy	σ acc

--

--

--

Per beat-type aggregate

Per scenario