Slot: settle_seed

Seed for the noise the settle integrates, when the settle runs as its own scan. Absent – the default – leaves the settle on whatever stream its solver hands it, which is what an unmodified upstream warm start does; the backend is not silently given a different draw than the reference implementation it is checked against. Present means the settle uses exactly this seed. Both interesting cases are then sayable: equal to seed puts the two scans on one stream, and different from it decorrelates them. Equal is not neutral – where the two scans take the same number of steps a shorter draw is a prefix of a longer one, so the measured noise repeats the settle’s sample for sample, and corr(settle, measured) on demeaned fluctuations is +0.49 against -0.01 decorrelated. Its effect on the reported statistics was null everywhere it was measured (FC bias below 0.03 sigma, no trend), so this slot exists to make the choice visible and reproducible rather than because either answer is known to be wrong.

URI: tvbo:slot/settle_seed

Applicable Classes

Name Description Modifies Slot
Noise no

Properties

Type and Range

Property Value
Range xsd:integer
Domain Of Noise

Cardinality and Requirements

Property Value

Slot Characteristics

Property Value
Owner Noise

Identifier and Mapping Information

Schema Source

  • from schema: https://w3id.org/tvbo

Mappings

Mapping Type Mapped Value
self tvbo:settle_seed
native tvbo:settle_seed

LinkML Source

name: settle_seed
description: 'Seed for the noise the settle integrates, when the settle runs as its
  own scan. Absent -- the default -- leaves the settle on whatever stream its solver
  hands it, which is what an unmodified upstream warm start does; the backend is not
  silently given a different draw than the reference implementation it is checked
  against. Present means the settle uses exactly this seed. Both interesting cases
  are then sayable: equal to `seed` puts the two scans on one stream, and different
  from it decorrelates them. Equal is not neutral -- where the two scans take the
  same number of steps a shorter draw is a prefix of a longer one, so the measured
  noise repeats the settle''s sample for sample, and corr(settle, measured) on demeaned
  fluctuations is +0.49 against -0.01 decorrelated. Its effect on the reported statistics
  was null everywhere it was measured (FC bias below 0.03 sigma, no trend), so this
  slot exists to make the choice visible and reproducible rather than because either
  answer is known to be wrong.'
from_schema: https://w3id.org/tvbo
rank: 1000
owner: Noise
domain_of:
- Noise
range: integer
required: false