Hi, I'm not really surprised that the grades are 'slightly chaotic' as you mention. Actually, I'd be surprised if a very simply heuristic like SM2 would be able to look inside your head and accurately predict when you're about to forget an item, given that it depends on so many external factors. I would need to analyse the data myself in more detail, but as I mentioned a few years before on this forum, I don't think it's worth making small tweaks to the scheduler, as that will probably not result in significantly different behaviour.
In summary, I just treat SM2 as a heuristic that periodically reminds me of things, without making claims that it's somehow optimal. With respect to your other remarks, for new cards SM2 is not yet followed: as soon as you grade it 2 or higher, the next rep will either be tomorrow or the day after depending on some randomness. Cheers, Peter -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Eric Sent: 18 April 2017 02:13 To: mnemosyne-proj-users <[email protected]> Subject: Re: [mnemosyne-proj-users] Is the interval-multiplier constant? בתאריך יום ראשון, 2 באפריל 2017 בשעה Hi, I'm back, and here is what I came up with. Regarding my question, I think it's important to point out that it actually means whether or not SM2 really works. The goal of the algorithm is to keep the knowledge at a useful level, so if it works well it should maintain at least a grade 4 level. And after looking in to it, I must say I can't really tell if it's working. Most objects never lift up to years-long spaces; people just neglect them after a short period of time. And apart from that, SM2 schedules repetitions by multiplying the previous space by the easiness factor, but Mnemosyne seems to have some other mysterious way. So basically, there isn’t enough data to work this out. Nevertheless, my general impression is that the data isn’t following SM2's basic assumption of a "factor by which the inter-repetition intervals should increase in successive repetitions"; moreover, the grades seem slightly chaotic to me. This is truly surprising. There are quite a few programs out there using this algorithm, and I had expected it to work at least in the short run, I was only skeptical about the long run. But there seems to be no evidence of it working at all. So I beg of anyone reading this, please share your experience, is this thing working or not? Having said that, here are some stuff I noticed which needs fixing: 1. The easiness factor increases upon a grade of 5, but it doesn’t take into account the scheduled interval, so when someone has a very early repetition and gets a 5, the program increases the EF even though there's no evidence it should be increased. 2. The scheduled intervals are measured in whole days, and the first is 0. So the new objects are usually being reviewed after several seconds or minutes, and if they pass the test they are scheduled for the next day, which is more than a 1000 times larger than the previous interval. Most of them fail this. Clearly, the logs show that the scheduled interval shouldn’t be measured in whole days. 3. The actual_interval is somewhat flawed, as it's mostly shorter than the interval between the timestamps, sometimes by a few seconds, and sometimes it's simply 0. 4. Sometimes an object has more than one 6-event, which is apparently the code for the event of adding a new object. All in all this was an eye-opening experience, and I thank you very much, Peter, for your generous help. 09:31:27 UTC+3, מאת Peter Bienstman: > Hi, > > If you want to look at your revision history in the database, the database is > a stardard sqlite database, which can be opened in many tools, e.g. > sqlitebrowser. > > Cheers, > > Peter > > > -----Original Message----- > > From: [email protected] [mailto:mnemosyne- > > [email protected]] On Behalf Of Eric > > Sent: 02 April 2017 01:04 > > To: mnemosyne-proj-users <[email protected]> > > Subject: Re: [mnemosyne-proj-users] Is the interval-multiplier constant? > > > > בתאריך יום שישי, 31 במרץ 2017 בשעה 14:30:53 UTC+3, מאת Peter Bienstman: > > > Hi, > > > > > > This is how the multiplication factor (card.easiness) changes as a > > > function of > > grade: > > > > > > if new_grade == 2: > > > card.easiness -= 0.16 > > > if new_grade == 3: > > > card.easiness -= 0.14 > > > if new_grade == 5: > > > card.easiness += 0.10 > > > if card.easiness < 1.3: > > > card.easiness = 1.3 > > > > > > Cheers, > > > > > > Peter > > > > > > -----Original Message----- > > > From: [email protected] [mailto:mnemosyne- > > [email protected]] On Behalf Of Eric > > > Sent: 31 March 2017 11:50 > > > To: mnemosyne-proj-users <[email protected]> > > > Subject: [mnemosyne-proj-users] Is the interval-multiplier constant? > > > > > > Hello, > > > I am trying to find out if the interval-multiplier is constant. > > > That is, if after a > > while you keep grading yourself 4? > > > I tried to check it in the DB that’s available on archive.org, but > > > MS Access > > couldn’t open the file. > > > So I was wondering if anyone can share his experience, or tell me > > > how to > > open the .db file. > > > Many thanks. > > > > > > -- > > > You received this message because you are subscribed to the Google > > Groups "mnemosyne-proj-users" group. > > > To unsubscribe from this group and stop receiving emails from it, > > > send an > > email to [email protected]. > > > To post to this group, send email to mnemosyne-proj- > > [email protected]. > > > To view this discussion on the web visit > > https://groups.google.com/d/msgid/mnemosyne-proj-users/7078c50e-f957 > > - 4f70-a09e-ef48ebbb08d1%40googlegroups.com. > > > For more options, visit https://groups.google.com/d/optout. > > > > Hi peter, thanks for your answer. > > > > I guess I wasn’t clear enough. I'm trying to find out if the factor > > is constant in our brain, not in the software. Actually what > > interests me most is whether it decreases. > > The software might take some time to adjust a card its true factor, > > so at first I guess the grades won't tell us much. But after a > > while, if the factor isn't decreasing in our brain, then the grades should > > look something like: > > 4,4,4,5,4,4,4,5,4,4,4. But if it is, then the grades should look something > > like: > > 4,4,4,3,4,4,4,3,4,4,4. > > > > As I mentioned before, I tried to check it in the database that’s > > available on archive.org, but MS Access couldn’t open the file. > > So I'm wondering if anyone can share their experience on that > > subject, or instruct me as how to open the .db file. > > > > I would greatly appreciate your help. > > > > -- > > You received this message because you are subscribed to the Google > > Groups "mnemosyne-proj-users" group. > > To unsubscribe from this group and stop receiving emails from it, > > send an email to [email protected]. > > To post to this group, send email to mnemosyne-proj- > > [email protected]. > > To view this discussion on the web visit > > https://groups.google.com/d/msgid/mnemosyne-proj-users/8f8307b2- > > 8bd4-4000-b35b-fa13de76ec44%40googlegroups.com. > > For more options, visit https://groups.google.com/d/optout. -- You received this message because you are subscribed to the Google Groups "mnemosyne-proj-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/mnemosyne-proj-users/221b424e-2d2d-4aee-97e9-0ba807f2b7b6%40googlegroups.com. For more options, visit https://groups.google.com/d/optout. -- You received this message because you are subscribed to the Google Groups "mnemosyne-proj-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/mnemosyne-proj-users/c6d238d208884204b07b1be5564e9bf4%40xmail101.UGent.be. For more options, visit https://groups.google.com/d/optout.
