Struct PredictiveModel 

pub struct PredictiveModel { /* private fields */ }

A predictive model that learns state transitions and rewards from experience.

This model enables an agent to forecast future states, predict rewards, and detect anomalies in its observations.

Implementations

impl PredictiveModel

pub fn new(config: PredictiveConfig) -> Self

Creates a new PredictiveModel with the given configuration.

pub fn with_default_config() -> Self

Creates a PredictiveModel with a default configuration.

pub fn record(&mut self, obs: &Observation)

Records an observation, updating the state history and anomaly detector.

pub fn record_transition( &mut self, obs: &Observation, action: &Action, reward: f64, next_obs: &Observation, )

Records a full state transition (s, a, r, s').

pub fn predict_next( &self, current: &Observation, action: &Action, ) -> PredictedState

Predicts the next state given the current state and an action.

pub fn predict_trajectory( &self, start: &Observation, actions: &[Action], ) -> Trajectory

Predicts a trajectory of future states for a given sequence of actions.

pub fn predict_reward(&self, state: &Observation, action: &Action) -> f64

Predicts the expected reward for taking a given action in a given state.

pub fn is_anomaly(&self, obs: &Observation) -> bool

Returns true if the observation is considered anomalous based on historical data.

pub fn get_confidence(&self, prediction: &PredictedState) -> f64

Returns the confidence score of a PredictedState.

pub fn get_uncertainty(&self, state: &Observation) -> f64

Returns an estimate of the model’s uncertainty about a given state. This uses the anomaly score as a proxy for uncertainty.

pub fn learn(&mut self)

Triggers the learning process for the model’s internal components.

pub fn history(&self) -> &VecDeque<StateSnapshot>

Returns a reference to the recent history of state snapshots.