AutoSF
- class AutoSF(embedding_dim=256, num_components=4, coefficients=[(0, 0, 0, 1), (1, 1, 1, 1), (2, 2, 2, 1), (3, 3, 3, 1), (1, 2, 3, - 1), (3, 1, 1, - 1)], embedding_kwargs=None, **kwargs)[source]
Bases:
pykeen.models.nbase.ERModel
An implementation of AutoSF from [zhang2020].
Initialize AutoSF via the
pykeen.nn.modules.AutoSFInteraction
interaction.Note
this variant uses num_components entity and relation representations with shared configuration. The coefficients should only be in \([0, num_components)\).
- Parameters
embedding_dim (
int
) – The entity embedding dimension \(d\).num_components (
int
) – the number of components.coefficients (
Sequence
[Tuple
[int
,int
,int
,Literal
[-1, 1]]]) – the coefficients determining the structure. The coefficients describe which head/relation/tail component get combined with each other. While in theory, we can have up to num_components**3 unique triples, usually, a smaller number is preferable to have some sparsity.embedding_kwargs (
Optional
[Mapping
[str
,Any
]]) – keyword arguments passed to the entity representationkwargs – Remaining keyword arguments passed through to
pykeen.models.ERModel
.
Attributes Summary
The default strategy for optimizing the model's hyper-parameters
Attributes Documentation