← Back to API reference

neuraltrain.models.linear.Linear

pydantic model neuraltrain.models.linear.Linear

Simple linear projection, with optional per-subject weights.

Parameters:

• reduction ({"mean", "concat"}) – How to reduce the time dimension before the linear layer. "mean" averages over time; "concat" flattens channels and time.

• subject_layers_config (SubjectLayers or None) – If set, use a SubjectLayersModel instead of a shared nn.Linear.

Fields:

• reduction (Literal['mean', 'concat'])

• subject_layers_config (neuraltrain.models.common.SubjectLayers | None)

field reduction: Literal['mean', 'concat'] = 'mean'

field subject_layers_config: SubjectLayers | None = None

build(n_in_channels: int, n_outputs: int) → Module

← Back to API reference