Crate pelt

Changepoint detection with Pruned Exact Linear Time

Changepoint detection using the PELT algorithm. This project is a port of the ruptures Python project. The PELT algorithm is described in this paper.

Example

extern crate pelt;
use std::fs;
use pelt::{MutEstimator, Pelt};

fn main() {
   let s = fs::read_to_string("./pelt-rs/signal.txt").unwrap();
   let signal: Vec<Vec<f64>> = vec!(s.split("\n")
   .map(|a| a.parse()
   .unwrap())
   .collect());

   let jump = Some(5);
   let min_size = Some(2);
   let loss = Some("l1");
   let pen = 3.;

   let mut m = Pelt::new(jump, min_size, loss, pen);
   println!("{:?}", m.predict(&signal))
}

Nightly

A nightly release is required for this crate.

Panics

This crate panics when the given signal length is smaller than the min_size field given to alg::Pelt.

Structs

Pelt	Pruned Exact Linear Time changepoint detection model.

Traits

MutEstimator

Trait mimicking Scikit-Learn API

Type Definitions

Vec2d

Stacked signals