Struct Adwin 

pub struct Adwin { /* private fields */ }

ADWIN streaming change detector.

Maintains a variable-length window of recent observations using an exponential histogram. After each add call the algorithm checks whether any sub-window split shows a statistically significant mean difference (Hoeffding bound at confidence 1 - δ). If so it shrinks the window to the more recent portion and reports drift.

Parameters

• delta — confidence parameter (smaller = fewer false positives, slower reaction). Typical value: 0.002.

Implementations

impl Adwin

pub fn new(delta: f64) -> Self

Create a new ADWIN detector with the given confidence parameter.

Lower delta values make the detector less sensitive (fewer false positives). 0.002 is a common choice.

pub fn add(&mut self, value: f64) -> bool

Add a new observation and check for drift.

Returns true if drift is detected (the window was shrunk).

Non-finite values (NaN, ±Inf) are rejected and the method returns false without modifying the window, because they would silently poison total_sum and total_variance.

pub fn mean(&self) -> f64

Current window mean.

pub fn window_size(&self) -> usize

Current window size (number of observations).

pub fn std_dev(&self) -> f64

Current estimated standard deviation of the window.

pub fn reset(&mut self)

Reset the detector to its initial state.

Trait Implementations

impl Clone for Adwin

fn clone(&self) -> Adwin

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for Adwin

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.