Struct ConvergenceMonitor 

pub struct ConvergenceMonitor { /* private fields */ }

Monitors convergence of iterative algorithms such as belief propagation.

Tracks residuals, manages damping schedules, and detects convergence or divergence.

Implementations

impl ConvergenceMonitor

pub fn new( config: ConvergenceConfig, schedule: DampingSchedule, ) -> Result<Self, ConvergenceError>

Create a new convergence monitor with the given configuration and schedule.

pub fn with_default_config() -> Self

Create a monitor with default configuration and fixed damping.

pub fn record_iteration(&mut self, residual: f64) -> bool

Record a new iteration with its residual.

Returns true if the algorithm should continue iterating, false if converged, diverged, or max iterations reached.

pub fn current_damping(&self) -> f64

Get the current damping factor.

pub fn state(&self) -> &ConvergenceState

Get a reference to the current convergence state.

pub fn is_converged(&self) -> bool

Check if the algorithm has converged.

pub fn is_diverged(&self) -> bool

Check if the algorithm has diverged.

pub fn iteration(&self) -> usize

Get the current iteration count.

pub fn reset(&mut self)

Reset the monitor to its initial state.

pub fn stats(&self) -> InferenceStats

Get summary statistics from the inference run.