duncan posted: "Another thing I am wondering about is the way that the grading affects scheduling- if I have a card that hasn't been reviewed for 7 days and I mark it a 2 it gets scheduled for 15 days in the future... "
I suspect that there will always be confusion about what the grades do in this type of algorithm. What the grades do is explained elsewhere. Here, I will try to explain *why it is confusing*, and what I've done about it. Part 1: Two Processes ******************** Two processes are ongoing: a. graduated intervals. Of course. This is the heart, without question. b. accumulating an easiness factor for each card. This is an attempt, for each card, to converge upon and remember an ef such that current_interval*ef is about right for that card. There is no doubt that this is a detail of secondary importance compared to a. Yet, this type of algorithm brings out the details of *b* to the user in the form of the buttons 2-5, which permit the increment/decrement of the ef, but do not have a decisive effect on what the *very next* interval will be. For example, citing your example above: grade 2: (2 - .16) * 7 (rounds down to 12) is very little different from grade 5: (2+.10) * 7 (rounds down to 14). (And also, you wanted something closer to 7, right?) The shorthand description of this behavior is this- For any passing grade, the algorithm stretches the interval by a factor around 2, and (secondarily) gradually adjusts this factor for each card. ((One key to resolving the persistent confusion about the grades would be to stress that that first function, the multiplication by "about 2" is the basic fact of operation, and that grades 2 and 5 have nowhere near the authority to override it.)) But, in view of the overriding centrality of *a*, I think that what you (duncan) wanted to do, and what many people evidently think should happen, is something like this: Grade 2: Almost forgot. Decrement ef, whatever, but *this time* don't increase the interval! I need to keep control of this item! new_interval=current_interval*(1 or very nearly 1) Grade 5: Let's fling this easy card out of the way for a while. Increment the ef, whatever, but *this time*, new_interval=current_interval*(some number noticeably larger than ef) So, for most users, this choice, 2-5, which seems momentous with regard to the interval, isn't. It only results in an adaptation to the user and/or the material studied, over some long period of usage. Part 2: A fundamental decision about forgotten cards *************************************************************** Choose one: a. a failed card has its memory-stretching sequence restarted at the beginning. b. a failed or almost failed card can be "held": I barely remember it for seven days, or barely forget it. Let me "hold" that; let me reduce or keep this interval, *this time*. Let me preserve the progress I've made on this item. Mnemosyne is type a of course. Again, this basic behavior dominates: For any passing grade, the interval increases by a factor around 2. If the card is failed, it starts over. This is very easy to understand! Part 3: What I have done to my personal copy of mnemosyne ******************************************************** I'm not a programmer, so this is only for myself. (I've disabled the data upload, in case that matters, because my data must not be valid, since I've changed the algorithm.) If it is desired that the user have the control over the interval (that he thinks he has) while also retaining the long-term adaptability made possible by the ef, *in a way that would then be transparent to the casual user*, something like this would work. The ef would be calculated as usual, but the intervals would be (something like this): Grade 2: new_interval=current_interval * 1.0 (hold) Grade 3: new_interval=current_interval * (ef * .75) (noticeably less) Grade 4: new_interval=current_interval * ef (<<raison d'etre of ef) Grade 5: new_interval=current_interval * (ef * 1.5) (noticeably more) See, I have a lot of authority to separate the cards quickly. And if this matters to you (for theoretical reasons), the median grade should approach 4 more quickly. Isn't that the goal of the ef idea? This was mysteriously difficult to figure out. I think Dr. SM just didn't think this through... :-) It is still monotonically increasing, ratcheting forward... Conclusion: ********** The reason for the persistent confusion is the need to explain to new users why the buttons don't do what they expect. The above makes them act as expected, while retaining the logic of the ef, which stays behind the scene for most users. --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "mnemosyne-proj-users" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/mnemosyne-proj-users?hl=en -~----------~----~----~----~------~----~------~--~---
