舍得英语魔法学苑

 找回密码
 注册
查看: 6872|回复: 0

[其它] Anki所用的算法版本

  [复制链接]
  • TA的每日心情
    开心
    2011-10-10 00:27
  • 签到天数: 68 天

    [LV.6]常住居民II

    舍得 发表于 2011-4-5 16:55:46 | 显示全部楼层 |阅读模式
    这是ANKI官方的回答。具体的链接如下:
    http://ankisrs.net/docs/FrequentlyAskedQuestions.html#_what_spaced_repetition_algorithm_does_anki_use

    Anki was originally based on the SuperMemo SM5 algorithm. However, Anki’sdefault behaviour of revealing the next interval before answering a cardrevealed some fundamental problems with the SM5 algorithm. The key differencebetween SM2 and later revisions of the algorithm is this:

    • SM2 uses your performance on a card to determine the next time to schedule   that card
    • SM3+ use your performance on a card to determine the next time to schedule   that card, and similar cards


    The latter approach promises to choose more accurate intervals by factoring innot just a single card’s performance, but the performance as a group. If youare very consistent in your studies and all cards are of a very similardifficulty, this approach can work quite well. However, once inconsistenciesare introduced into the equation (cards of varying difficulty, not studying atthe same time every day), SM3+ is more prone to incorrect guesses at the nextinterval - resulting in cards being scheduled too often or too far in thefuture.

    Furthermore, as SM3+ dynamically adjusts the \"optimum factors\" table, asituation can often arise where answering \"hard\" on a card can result in alonger interval than answering \"easy\" would give. The next times are hiddenfrom you in SuperMemo so the user is never aware of this.

    After evaluating the alternatives, the Anki author decided that near-optimumintervals yielded by an SM2 derivative are better than trying to obtainoptimum intervals at the risk of incorrect guesses. An SM2 approach ispredictable and intuitive to end users, whereas an SM3+ approach hides thedetails from the user and requires users to trust the system (even when thesystem may make mistakes in the scheduling).

    Anki’s algorithm is based on SM2, but differs from it in some respects.

    • SM2 defines an initial interval of 1 day then 6 days. With Anki, you have   the choice of 1, 3-5 and 7-9 days depending on your initial answer. After   that, the intervals are increased in largely the same way. Anki understands   that it can be necessary to see a new card a number of times before you’re   able to memorize it, and those initial failures don’t mean you need to be   punished by being shown the failed card many times over the course of a   few days. Performance during the learning stage does not reflect   performance in the retaining stage.
    • Anki uses 4 choices for answering, not 6. There is only one fail choice,   not 3. The reason for this is that failure comprises a small amount of   total reviews, and thus adjusting a card’s ease can be sufficiently done by   simply varying the positive answers.
    • Answering cards later than scheduled will be factored into the next   interval calculation, so you receive a boost to cards that you were late in   answering but still remembered.
    • SM2 does not define a specific order to show the cards in. Anki has support   for different priorities, and the user can choose the order that cards due   for review are displayed in. The default order shows oldest cards first,   which makes the first half of a review session easier and reduces the   chances of a user falling behind in their reviews and becoming demotivated.
    • Like SM2, Anki’s failure button resets the card interval by default. But   the user can choose to make the card take some steps back rather than being   reset completely. Also, you can elect to review failed mature cards on a   different day, instead of the same day.
    • Learning new material is integrated into the review process. Failing cards   results in them being presented again for review within 10 minutes by   default, instead of all at the end like SuperMemo’s   final drill option. There is also an option to keep failed cards to the   end.
    • Remembered easily not only increments the ease factor, but adds an extra   bonus to the current interval calculation. Thus, answering remembered   easily is a little more aggressive than the standard SM2 algorithm.
    • Successive failures while cards are young (under 21 days) do not result   in further decreases to the card’s ease. A common complaint with the   standard SM algorithms is that repeated failings of a card cause the card   to get stuck in \"low interval hell\". In Anki, the initial acquisition   process does not influence a card’s ease as much as it does in   SuperMemo.
    • Ease factors start off at 2.5 for the very first new card, but subsequent   new cards are based on the average ease factor of the deck. Thus in a deck   where the average easiness is very low, new cards will start out with more   conservative intervals. Note that once a new card has been answered once,   future scheduling is independent of other cards and thus its progress is   predictable, unlike SM3+.


    For a description of the SM2 algorithm, please seehttp://www.supermemo.com/english/ol/sm2.htm




    该贴已经同步到 舍得的微博
    您需要登录后才可以回帖 登录 | 注册

    本版积分规则

    小黑屋|手机版|Archiver|官方微博|官方QQ群|舍得英语魔法学苑 ( 冀ICP备11024081号-1 )

    GMT+8, 2018-6-22 00:21 , Processed in 0.072466 second(s), 21 queries .

    Powered by Discuz! X3.2

    © 2001-2013 Comsenz Inc.

    快速回复 返回顶部 返回列表