1 unstable release

0.1.0 Mar 5, 2022

#15 in #prediction

MIT license

9KB
174 lines

Context Tree Weighting (CTW)

CTW is a lightweight, practical and well performing sequence prediction algorithm discovered by Frans Willems, Yuri Shtarkov and Tjalling Tjalkens in 1995.

It has good query and update performance (linear in the context length).

Useage

The following example demonstrates Ctw learning the altenating binary sequence 101010...

use ctw::CtwTree;
let mut rng = rand::thread_rng();
let mut tree = CtwTree::new(8); // context length is 8

let pattern = [true, false].into_iter().cycle().take(100); // true, false, true, ..
tree.update_batch(&Vec::from_iter(pattern));

let mut predictions = Vec::new();

for _ in 0..10 {
    let prediction = tree.sample(&mut rng);
    tree.update(prediction);

    // convert bool to 1 and 0 for legibility
    let bit = if prediction { 1 } else { 0 };
    predictions.push(bit);
}

println!("predictions: {predictions:?}"); // --> "prediction: [1, 0, 1, 0, 1, 0, 1, 0, 1, 0]"
assert_eq!(predictions, vec![1,0,1,0,1,0,1,0,1,0]);

Resources

  1. http://www.hutter1.net/publ/sctw.pdf
  2. https://web.stanford.edu/class/ee378a/lecture-notes/lecture_9.pdf
  3. http://www.data-compression.info/Algorithms/CTW/
  4. https://www.cs.cmu.edu/~aarti/Class/10704_Spring15/CTW.pdf (original paper)

Dependencies

~315KB