pub trait HasReadout: StreamingLearner {
// Required method
fn readout_weights(&self) -> &[f64];
}Expand description
Models that expose a linear readout weight vector.
Implemented by models with an RLS readout layer: neural models (ESN, Mamba,
KAN, TTT, sLSTM, HGRN2, mGRADE, attention variants), kernel models (KRLS),
and linear models (RLS). Used by ProjectedLearner for supervised
projection updates.
§Object Safety
This trait is object-safe. Box<dyn HasReadout> is a legal type.
Required Methods§
Sourcefn readout_weights(&self) -> &[f64]
fn readout_weights(&self) -> &[f64]
The linear readout weight vector.
For RLS-family models this is the full coefficient vector including the bias term if one is used. Length matches the model’s internal feature dimensionality.