salmon.triplets.samplers.Adaptive
- class salmon.triplets.samplers.Adaptive(*, n: int, d: int = 2, ident: str = '', module: str = 'TSTE', optimizer: str = 'Embedding', R: float = 10, scorer: str = 'infogain', random_state: Optional[int] = None, **kwargs)
The sampler that runs adaptive algorithms.
- __init__(*, n: int, d: int = 2, ident: str = '', module: str = 'TSTE', optimizer: str = 'Embedding', R: float = 10, scorer: str = 'infogain', random_state: Optional[int] = None, **kwargs)
- Parameters
- nint
The number of items to embed.
- dint (optional, default:
2) Embedding dimension.
- identstr (optional, default:
""). The identity of this runner. Must be unique among all adaptive algorithms.
- optimizerstr (optional, default:
Embedding). The optimizer underlying the embedding. This method specifies how to change the batch size. Choices are
["Embedding", "PadaDampG", "GeoDamp"].- Rint (optional, default:
1) Adaptive sampling after
R * nresponses have been received.- scorerstr (optional, default:
"infogain") How queries should be scored. Scoring with
scorer='infogain'tries to link query score and “embedding improvement,” andscorer='uncertainty'looks at the query that’s closest to the decision boundary (or 50% probability).- random_stateint, None, optional (default:
None) The random state to be used for initialization.
- kwargsdict, optional
Keyword arguments to pass to
Embedding.