neuraltrain.models.fmri_mlp.FmriLinear¶

class neuraltrain.models.fmri_mlp.FmriLinear(*, out_dim: int | None = None, time_agg: Literal['in_mean', 'in_linear', 'out_mean', 'out_linear'] = 'out_linear', output_head_config: Mlp | dict[str, Mlp] | None = None)[source][source]¶

Single linear layer for fMRI decoding with temporal aggregation.

Parameters:

out_dim (int or None) – Explicit output dimension. Overridden by n_outputs at build time.
time_agg ({"in_mean", "in_linear", "out_mean", "out_linear"}) – Temporal aggregation strategy. "in_*" aggregates before the linear layer; "out_*" aggregates after.
output_head_config (Mlp, dict[str, Mlp], or None) – Optional output head(s) appended after the linear layer.