Struct StreamingMamba 

pub struct StreamingMamba { /* private fields */ }

Streaming Mamba model implementing StreamingLearner.

Combines a selective SSM for temporal feature extraction with an RLS readout layer. The SSM processes each input as a timestep, evolving hidden state to capture temporal dependencies. The RLS layer learns a linear mapping from SSM outputs to the regression target.

Example

use irithyll::ssm::{StreamingMamba, MambaConfig};
use irithyll::learner::StreamingLearner;

let config = MambaConfig::builder()
    .d_in(3)
    .n_state(8)
    .build()
    .unwrap();

let mut model = StreamingMamba::new(config);

// Train on a stream of 3-dimensional features
for i in 0..100 {
    let x = [i as f64 * 0.1, (i as f64).sin(), 1.0];
    let y = x[0] + 0.5 * x[1];
    model.train(&x, y);
}

let pred = model.predict(&[10.0, 0.0, 1.0]);
assert!(pred.is_finite());

Implementations

impl StreamingMamba

pub fn new(config: MambaConfig) -> Self

Create a new streaming Mamba model from the given configuration.

Initializes the SSM with random weights (seeded by config.seed) and an RLS readout with the specified forgetting factor and P matrix scale.

pub fn config(&self) -> &MambaConfig

Get a reference to the model configuration.

pub fn ssm_state(&self) -> &[f64]

Get the current SSM hidden state.

pub fn last_features(&self) -> &[f64]

Get the cached SSM output features from the last training step.

Trait Implementations

impl StreamingLearner for StreamingMamba

fn train_one(&mut self, features: &[f64], target: f64, weight: f64)

Train on a single observation with explicit sample weight.

fn predict(&self, features: &[f64]) -> f64

Predict the target for the given feature vector.

fn n_samples_seen(&self) -> u64

Total number of observations trained on since creation or last reset.

fn reset(&mut self)

Reset the model to its initial (untrained) state.

fn train(&mut self, features: &[f64], target: f64)

Train on a single observation with unit weight.

fn predict_batch(&self, feature_matrix: &[&[f64]]) -> Vec<f64>

Predict for each row in a feature matrix.