Struct Engine 

pub struct Engine<M: DistanceMetric> { /* private fields */ }

High-level facade for matrix profile computation, generic over distance metric.

Examples

use motif_rs::{EuclideanEngine, MatrixProfileConfig};

let ts = vec![1.0, 2.0, 3.0, 2.0, 1.0, 2.0, 3.0, 2.0, 1.0];
let engine = EuclideanEngine::new(MatrixProfileConfig::new(4));
let mp = engine.compute(&ts);
assert_eq!(mp.profile.len(), ts.len() - 4 + 1);

Implementations

impl<M: DistanceMetric> Engine<M>

pub fn new(config: MatrixProfileConfig) -> Self

Create a new engine with the given configuration.

pub fn compute(&self, ts: &[f64]) -> MatrixProfile

Compute the full matrix profile for a time series (batch STOMP).

pub fn streaming(&self, initial_ts: &[f64], egress: bool) -> Stampi<M>

Create a streaming matrix profile from an initial time series.

• egress=false: grow mode — time series extends unboundedly.
• egress=true: egress mode — fixed-size sliding window.

pub fn ab_join(&self, ts_a: &[f64], ts_b: &[f64]) -> (JoinProfile, JoinProfile)

Compute the AB-join between two time series.

Returns two JoinProfiles: one for each series against the other.

pub fn compute_topk(&self, ts: &[f64], k: usize) -> TopKMatrixProfile

Compute the top-k matrix profile for a time series.

Stores the k nearest neighbors for each subsequence, rather than just the best one.

pub fn snippets(&self, ts: &[f64], k: usize) -> SnippetsResult

Extract k representative snippets that best summarize the time series.

Uses z-normalized Euclidean distance profiles regardless of the engine’s metric.

pub fn mpdist(&self, ts_a: &[f64], ts_b: &[f64], percentage: Option<f64>) -> f64

Compute MPdist: a scalar distance between two time series.

Based on the k-th percentile of the concatenated AB-join profiles.

pub fn scrump(&self, ts: &[f64], percentage: f64) -> MatrixProfile

Compute an approximate matrix profile using SCRUMP/PreSCRIMP.

percentage in (0.0, 1.0] controls the fraction of diagonals sampled. At 1.0, delegates to exact STOMP.

pub fn ostinato(&self, ts_list: &[&[f64]]) -> ConsensusMotif

Find the consensus motif across multiple time series.

Returns the subsequence (from any series) whose maximum nearest-neighbor distance to all other series is minimized.

pub fn stimp( &self, ts: &[f64], min_m: usize, max_m: usize, step: Option<usize>, percentage: Option<f64>, ) -> PanMatrixProfile

Compute the pan matrix profile across a range of window sizes.

Profiles are normalized by 1/sqrt(2*m) for cross-window comparability.

pub fn compute_pnorm(&self, ts: &[f64], p: f64) -> MatrixProfile

Compute the matrix profile using Minkowski p-norm distance.

For p == 2.0, delegates to the optimized AAMP path. For other values of p, uses a diagonal recurrence. This is AAMP-only (non-normalized); the engine’s metric type M is ignored.

pub fn ab_join_pnorm( &self, ts_a: &[f64], ts_b: &[f64], p: f64, ) -> (JoinProfile, JoinProfile)

Compute the AB-join between two time series using Minkowski p-norm distance.

Returns two JoinProfiles, one for each series. This is AAMP-only (non-normalized); the engine’s metric type M is ignored.

pub fn mpdist_pnorm( &self, ts_a: &[f64], ts_b: &[f64], p: f64, percentage: Option<f64>, ) -> f64

Compute MPdist using Minkowski p-norm distance.

This is AAMP-only (non-normalized); the engine’s metric type M is ignored.