Struct NeuralMoE 

pub struct NeuralMoE { /* private fields */ }

Streaming Neural Mixture of Experts.

Polymorphic MoE where each expert can be any StreamingLearner. Implements StreamingLearner itself for composability.

Example

use irithyll::moe::NeuralMoE;
use irithyll::{sgbt, esn, StreamingLearner};

let mut moe = NeuralMoE::builder()
    .expert(sgbt(50, 0.01))
    .expert(sgbt(100, 0.005))
    .expert_with_warmup(esn(50, 0.9), 50)
    .top_k(2)
    .build();

moe.train(&[1.0, 2.0, 3.0], 4.0);
let pred = moe.predict(&[1.0, 2.0, 3.0]);

Implementations

impl NeuralMoE

pub fn builder() -> NeuralMoEBuilder

Start building a NeuralMoE with the builder pattern.

impl NeuralMoE

pub fn n_experts(&self) -> usize

Number of experts.

pub fn top_k(&self) -> usize

Current top-k setting.

pub fn utilization(&self) -> Vec<f64>

Per-expert utilization (EWMA of routing probability).

pub fn expert_samples(&self) -> Vec<u64>

Per-expert samples trained.

pub fn n_dead_experts(&self) -> usize

Number of dead experts (utilization below threshold).

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

Load distribution from the router.

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

Expert disagreement: std dev of all expert predictions.

High disagreement indicates the experts have divergent views on this input — a real uncertainty signal. Returns 0.0 when fewer than 2 experts exist or predictions are uniform.

pub fn cached_disagreement(&self) -> f64

Cached expert disagreement from the most recent train_one call.

This avoids recomputing disagreement in config_diagnostics(), which only has &self (no features available). Updated every train_one.

pub fn expert_predictions(&self, features: &[f64]) -> Vec<f64>

Get predictions from all experts (for inspection).

pub fn routing_probabilities(&self, features: &[f64]) -> Vec<f64>

Get current routing probabilities (for inspection).

Trait Implementations

impl DiagnosticSource for NeuralMoE

fn config_diagnostics(&self) -> Option<ConfigDiagnostics>

Return config diagnostics, or None if not supported.

impl StreamingLearner for NeuralMoE

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 diagnostics_array(&self) -> [f64; 5]

Raw diagnostic signals for adaptive tuning.

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.

fn adjust_config(&mut self, _lr_multiplier: f64, _lambda_delta: f64)

Apply smooth learning rate and regularization adjustments.

fn apply_structural_change(&mut self, _depth_delta: i32, _steps_delta: i32)

Apply structural changes at model replacement boundaries.

fn replacement_count(&self) -> u64

Total number of internal model replacements (e.g. tree replacements triggered by drift detection or max-tree-samples).

fn check_proactive_prune(&mut self) -> bool

Manually trigger a proactive prune check.

fn set_prune_half_life(&mut self, _hl: usize)

Dynamically set the contribution accuracy EWMA half-life.

fn readout_weights(&self) -> Option<&[f64]>

Return the readout weight vector for supervised projection, if available.

fn tree_structure(&self) -> Vec<(usize, usize, f64, f64, u64)>

Optional tree-level structure diagnostics.