Hacker News new | ask | show | jobs
by adaszko 1273 days ago
There’s a Bayesian stats approach to spaced repetition: https://fasiha.github.io/ebisu/

AFAIU, SM2 computes the datetime of a next review, whereas Ebisu models a probability of remembering a given flashcard. It seems it’s a more straightforward representation that’s more amenable to implementing functionalities like “show me 10 least remembered cards”.
1 comments
cyphar 1273 days ago
It should be noted that Ebisu currently has a severe flaw that causes it to massively underestimate how well you know a card[1]. The author has been working on a rewrite to fix the issue for a year and a half, and is currently asking for comments on the new API for Ebisu v3 which should fix this issue[2]. (The author argues this never would've caused to you to get suboptimal reviews if you use it in the "just give me a few reviews now" mode, but I'm not personally convinced.)

All of that being said, I am a fan of the Bayesian approach Ebisu uses (and Ebisu v3 would allow you to eliminate the awful way SM2/Anki deal with ease factors). Thankfully the v3 Anki scheduler will be pluggable so you'll be able to use Ebisu v3 with Anki on every Anki platform. :D

[1]: https://github.com/fasiha/ebisu/issues/43 [2]: https://github.com/fasiha/ebisu/issues/58

link
adaszko 1273 days ago
Oh, interesting. Is v3 just a new API or a new algorithm altogether?
link
cyphar 1267 days ago
It's still Anki's version of SM-2, with minor changes to its behaviour. You can see a list of the changes here[1].

[1]: https://faqs.ankiweb.net/the-2021-scheduler.html

link