Learning Algorithms¶

A learning algorithm is a formula that determines when a note or flashcard should next be reviewed.

SM-2-OSR¶

The SM-2-OSR algorithm is a variant of Anki's algorithm which is based on the SM-2 algorithm.
It supports ternary reviews i.e. a concept is either hard, good, or easy at the time of review.
initial ease is weighted (using max_link_factor) depending on the average ease of linked notes, note importance, and the base ease.
Anki also applies a small amount of random “fuzz” to prevent cards that were introduced at the same time and given the same ratings from sticking together and always coming up for review on the same day."
The algorithm is essentially the same for both notes and flashcards - apart from the PageRanks

Warning

Note that this hasn't been updated in a while, please see the code.

if link_count > 0: initial_ease = (1 - link_contribution) * base_ease + link_contribution * average_ease - link_contribution = max_link_factor * min(1.0, log(link_count + 0.5) / log(64)) (cater for uncertainty)
- The importance of the different concepts/notes is determined using the PageRank algorithm (not all notes are created equal xD)
  - On most occasions, the most fundamental concepts/notes have higher importance
If the user reviews a concept/note as:
- easy, the ease increases by 20 and the interval changes to old_interval * new_ease / 100 * 1.3 (the 1.3 is the easy bonus)
- good, the ease remains unchanged and the interval changes to old_interval * old_ease / 100
- hard, the ease decreases by 20 and the interval changes to old_interval * 0.5
  - The 0.5 can be modified in settings
  - minimum ease = 130
- For 8 or more days:
  - interval += random_choice({-fuzz, 0, +fuzz})
    - where fuzz = ceil(0.05 * interval)
    - Anki docs: > "[...] Anki also applies a small amount of random “fuzz” to prevent cards that were introduced at the same time and given the same ratings from sticking together and always coming up for review on the same day."
The scheduling information is stored in the YAML front matter

The algorithm is detailed at: fsrs4anki

Incorporation of the FSRS algorithm into this plugin has not yet occurred. For progress see: [FEAT] sm-2 is outdated, can you please replace it with the fsrs algorithm? #748

This is the simplest "algorithm" possible. There are fixed intervals configured by the user for each of the possible review outcomes.

For example, hard might be configured for an interval of 1 day.